Settings reload: no DNS at build time, don't refuse over old failures

A 3-node rehearsal found every settings reload refused, cluster-wide,
because one node couldn't resolve the Pyzor server:

- PyzorConfig::parse resolved the host while building the settings and
  made a failed lookup a build error. It now keeps the host and port and
  resolves when a message is checked (an IP address is used as is, a
  name is reused for five minutes, the lookup counts against the Pyzor
  timeout). A failure there is a Pyzor error for that message.
- A milter's hostname was resolved the same way, with a blocking
  to_socket_addrs in async code. An IP address is kept; a name is now
  resolved on each connection.

Other build-time I/O is already non-fatal: directories that can't
connect become unavailable with a warning (DIR-21), and the AI model
locality check only warns.

reload_registry swapped the core only when the whole build was free of
errors, while boot runs with whatever built. One failing object thus
refused every later reload, and the running settings went stale. Now a
reload is refused only for errors in objects that built when the
running settings were built (at boot or by the last applied reload):
applying it would lose those. Objects that already failed then are
missing from the running settings anyway, as at boot, so their errors
are logged and returned as known_errors but don't hold the reload back.
Refusing on new errors keeps a bad edit from taking a working object
out of service; the admin gets the error instead.

ReloadSettings now says "Settings were not reloaded." and names the
object and its error ("Tracer with id ...: Only one console tracer is
allowed"), with a count of any further errors. A refused reload after a
directory change logs its errors too.

system::reload::reload_tests (new): with Pyzor enabled on an
unresolvable host, ReloadSettings succeeds (on main it fails with
"Invalid address: failed to lookup address information"); an IP host
needs no lookup; a new build error refuses the reload, names the object
and leaves the running settings unchanged; the same error, once known
from the running settings' build, no longer blocks; once fixed, a new
error there blocks again. smtp::inbound::milter's session test now
names its milter "localhost", so the connect-time lookup is exercised.

crates/common/src/cache/reload.rs

@@ -20,13 +20,20 @@ use ahash::AHashMap;
 use directory::Directories;
 use registry::{
     schema::{prelude::ObjectType, structs::BlockedIp},
-    types::error::{Error, Warning},
+    types::{
+        error::{Error, Warning},
+        id::ObjectId,
+    },
 };
 use std::sync::Arc;
 use store::{LookupStores, registry::bootstrap::Bootstrap, write::now};
 
 pub struct ReloadResult {
+    /// Errors that kept the reload from being applied.
     pub errors: Vec<Error>,
+    /// inbuxa: errors in objects that already failed when the running
+    /// settings were built; logged, but they don't refuse a reload.
+    pub known_errors: Vec<Error>,
     pub warnings: Vec<Warning>,
     pub replaced_core: bool,
 }
@@ -114,24 +121,30 @@ impl Server {
                     directories: directory.directories,
                 };
 
-                // Parse tracers
+                // inbuxa: upstream swapped the core only when the whole build
+                // was free of errors, while boot runs with whatever built. So one
+                // object that failed (a DNS lookup that timed out, say) refused
+                // every later reload, cluster-wide when the reload came from
+                // ReloadSettings, and the running settings went stale. Now a
+                // reload is refused only for errors in objects that built when
+                // the running settings were built: those would be lost by
+                // applying it. Objects that already failed then are missing
+                // from the running settings anyway, as at boot, so their
+                // errors are reported but don't hold the reload back.
                 let tracers = Telemetry::parse(&mut bootstrap, &storage).await;
-
-                if bootstrap.errors.is_empty() {
                 let core = Box::pin(Core::parse(&mut bootstrap, storage)).await;
-
-                    if bootstrap.errors.is_empty() {
                 let mut servers = Listeners::parse(&mut bootstrap).await;
+
+                if !self.has_new_build_errors(&bootstrap.errors) {
                     servers
                         .parse_tcp_acceptors(&mut bootstrap, self.inner.clone())
                         .await;
 
-                        if bootstrap.errors.is_empty() {
+                    if !self.has_new_build_errors(&bootstrap.errors) {
                         // Update core
                         self.inner.shared_core.store(core.into());
 
                         // Update tracers
-
                         tracers.update();
 
                         // Reload queue settings
@@ -142,14 +155,26 @@ impl Server {
                             .await
                             .ok();
 
+                        self.record_build_errors(&bootstrap.errors);
+
                         return Ok(ReloadResult {
-                                errors: bootstrap.errors,
+                            errors: Vec::new(),
+                            known_errors: bootstrap.errors,
                             warnings: bootstrap.warnings,
                             replaced_core: true,
                         });
                     }
                 }
-                }
+
+                let (known_errors, errors) = std::mem::take(&mut bootstrap.errors)
+                    .into_iter()
+                    .partition(|error| self.is_known_build_error(error));
+                return Ok(ReloadResult {
+                    errors,
+                    known_errors,
+                    warnings: bootstrap.warnings,
+                    replaced_core: false,
+                });
             }
         }
 
@@ -163,7 +188,7 @@ impl ReloadResult {
     }
 
     pub fn log(&self) {
-        for error in &self.errors {
+        for error in self.errors.iter().chain(&self.known_errors) {
             error.log();
         }
         for warning in &self.warnings {
@@ -176,8 +201,34 @@ impl From<Bootstrap> for ReloadResult {
     fn from(bootstrap: Bootstrap) -> Self {
         Self {
             errors: bootstrap.errors,
+            known_errors: Vec::new(),
             warnings: bootstrap.warnings,
             replaced_core: false,
         }
     }
 }
+
+// inbuxa: which objects failed to build for the running settings
+impl Server {
+    /// Records the objects that failed to build for the settings now running.
+    pub fn record_build_errors(&self, errors: &[Error]) {
+        *self.inner.data.build_errors.lock() = errors.iter().filter_map(error_object).collect();
+    }
+
+    fn is_known_build_error(&self, error: &Error) -> bool {
+        error_object(error).is_some_and(|id| self.inner.data.build_errors.lock().contains(&id))
+    }
+
+    fn has_new_build_errors(&self, errors: &[Error]) -> bool {
+        errors.iter().any(|error| !self.is_known_build_error(error))
+    }
+}
+
+fn error_object(error: &Error) -> Option<ObjectId> {
+    match error {
+        Error::Validation { object_id, .. }
+        | Error::Build { object_id, .. }
+        | Error::NotFound { object_id } => Some(*object_id),
+        Error::Internal { object_id, .. } => *object_id,
+    }
+}

crates/common/src/config/inner.rs

@@ -96,6 +96,7 @@ impl Data {
             applications,
             logos: Default::default(),
             smtp_connectors: TlsConnectors::try_new().failed("Failed to build TLS connectors"),
+            build_errors: Default::default(),
             asn_geo_data: Default::default(),
         }
     }
@@ -237,6 +238,7 @@ impl Default for Data {
             applications: WebApplications::new(),
             logos: Default::default(),
             smtp_connectors: TlsConnectors::try_new().unwrap(),
+            build_errors: Default::default(),
             asn_geo_data: Default::default(),
             lookup_stores: Default::default(),
         }

crates/common/src/config/mailstore/spamfilter.rs

@@ -16,7 +16,6 @@ use mail_auth::common::resolver::ToReverseName;
 use nlp::classifier::model::{CcfhClassifier, FhClassifier};
 use registry::schema::{
     enums::{ExpressionVariable, ModelSize},
-    prelude::ObjectType,
     structs::{
         self, SpamDnsblServer, SpamDnsblSettings, SpamFileExtension, SpamPyzor, SpamRule,
         SpamSettings, SpamTag,
@@ -25,10 +24,10 @@ use registry::schema::{
 use sieve::SpamStatus;
 use std::{
     net::{IpAddr, SocketAddr},
-    time::Duration,
+    sync::Arc,
+    time::{Duration, Instant},
 };
 use store::registry::{RegistryObject, bootstrap::Bootstrap};
-use tokio::net::lookup_host;
 use utils::{cache::CacheItemWeight, glob::GlobMap};
 
 #[derive(rkyv::Archive, rkyv::Deserialize, rkyv::Serialize, Debug, Default)]
@@ -157,7 +156,11 @@ pub struct FtrlParameters {
 
 #[derive(Debug, Clone)]
 pub struct PyzorConfig {
-    pub address: SocketAddr,
+    // inbuxa: the server is resolved when a message is checked, not while the
+    // settings are built (see PyzorConfig::address)
+    pub host: String,
+    pub port: u16,
+    pub resolved: Arc<parking_lot::Mutex<Option<(SocketAddr, Instant)>>>,
     pub timeout: Duration,
     pub min_count: u64,
     pub min_wl_count: u64,
@@ -474,31 +477,15 @@ impl PyzorConfig {
             return None;
         }
 
-        let port = pyzor.port;
+        // inbuxa: upstream resolved the host here and reported a failed lookup
-        let host = pyzor.host;
+        // as a build error, so a DNS hiccup on one node refused every settings
-        let address = match lookup_host(format!("{host}:{port}"))
+        // reload on it (and, from the node that ran ReloadSettings, across the
-            .await
+        // cluster). The lookup now happens when a message is checked; a
-            .map(|mut a| a.next())
+        // failure there is logged as a Pyzor error for that message.
-        {
-            Ok(Some(address)) => address,
-            Ok(None) => {
-                bp.build_error(
-                    ObjectType::SpamPyzor.singleton(),
-                    "Invalid address: No addresses found.",
-                );
-                return None;
-            }
-            Err(err) => {
-                bp.build_error(
-                    ObjectType::SpamPyzor.singleton(),
-                    format!("Invalid address: {}", err),
-                );
-                return None;
-            }
-        };
-
         PyzorConfig {
-            address,
+            host: pyzor.host,
+            port: pyzor.port as u16,
+            resolved: Default::default(),
             timeout: pyzor.timeout.into_inner(),
             min_count: pyzor.block_count,
             min_wl_count: pyzor.allow_count,
@@ -508,6 +495,35 @@ impl PyzorConfig {
     }
 }
 
+// inbuxa: how long a resolved Pyzor address is reused
+const PYZOR_RESOLVE_TTL: Duration = Duration::from_secs(300);
+
+impl PyzorConfig {
+    /// The server's address: the host itself when it is an IP address,
+    /// otherwise the first address it resolves to, reused for five minutes.
+    pub async fn address(&self) -> std::io::Result<SocketAddr> {
+        if let Ok(ip) = self.host.parse::<IpAddr>() {
+            return Ok(SocketAddr::new(ip, self.port));
+        }
+        if let Some((address, resolved_at)) = *self.resolved.lock()
+            && resolved_at.elapsed() < PYZOR_RESOLVE_TTL
+        {
+            return Ok(address);
+        }
+        let address = tokio::net::lookup_host((self.host.as_str(), self.port))
+            .await?
+            .next()
+            .ok_or_else(|| {
+                std::io::Error::new(
+                    std::io::ErrorKind::NotFound,
+                    format!("{} has no addresses", self.host),
+                )
+            })?;
+        *self.resolved.lock() = Some((address, Instant::now()));
+        Ok(address)
+    }
+}
+
 impl ClassifierConfig {
     pub async fn parse(bp: &mut Bootstrap) -> Option<Self> {
         let classifier = bp.setting_infallible::<structs::SpamClassifier>().await;

crates/common/src/config/smtp/session.rs

@@ -2,6 +2,8 @@
  * SPDX-FileCopyrightText: 2020 Stalwart Labs LLC <[email protected]>
  *
  * SPDX-License-Identifier: AGPL-3.0-only OR LicenseRef-SEL
+ *
+ * Modified by Coffey Labs in 2026 for INBUXA.
  */
 
 use self::resolver::Policy;
@@ -22,7 +24,7 @@ use registry::schema::{
 };
 use smtp_proto::*;
 use std::{
-    net::{SocketAddr, ToSocketAddrs},
+    net::{IpAddr, SocketAddr},
     str::FromStr,
     time::Duration,
 };
@@ -384,19 +386,16 @@ impl SessionConfig {
                     Some(Milter {
                         enable: bp.compile_expr(id, &milter.ctx_enable()),
                         id,
-                        addrs: format!("{}:{}", milter.hostname, milter.port)
+                        // inbuxa: upstream resolved the hostname here (a
-                            .to_socket_addrs()
+                        // blocking lookup) and made a failure a build error,
-                            .map_err(|err| {
+                        // which refused the whole settings reload. An IP
-                                bp.build_error(
+                        // address is kept as is; a name is resolved on each
-                                    id,
+                        // connection (MilterClient::connect).
-                                    format!(
+                        addrs: milter
-                                        "Unable to resolve milter hostname {}: {}",
+                            .hostname
-                                        milter.hostname, err
+                            .parse::<IpAddr>()
-                                    ),
+                            .map(|ip| vec![SocketAddr::new(ip, milter.port as u16)])
-                                )
+                            .unwrap_or_default(),
-                            })
-                            .ok()?
-                            .collect(),
                         hostname: milter.hostname,
                         port: milter.port as u16,
                         timeout_connect: milter.timeout_connect.into_inner(),

crates/common/src/ipc.rs

@@ -335,3 +335,57 @@ impl EmailPush {
         }
     }
 }
+
+/// inbuxa: the task locks this node holds, so a graceful stop can hand them
+/// back instead of leaving the tasks blocked until the locks expire.
+pub struct TaskLocks {
+    held: parking_lot::Mutex<ahash::AHashSet<u64>>,
+    stopping: AtomicBool,
+    expiry: std::sync::atomic::AtomicU64,
+}
+
+impl TaskLocks {
+    /// How long a task lock lasts, in seconds, unless it is released first.
+    pub const DEFAULT_EXPIRY: u64 = 60 * 60;
+
+    pub fn is_stopping(&self) -> bool {
+        self.stopping.load(Ordering::Acquire)
+    }
+
+    /// Stops new claims and returns the ids of every lock still held.
+    pub fn stop(&self) -> Vec<u64> {
+        self.stopping.store(true, Ordering::Release);
+        self.held.lock().drain().collect()
+    }
+
+    pub fn insert(&self, id: u64) {
+        self.held.lock().insert(id);
+    }
+
+    pub fn remove(&self, id: u64) {
+        self.held.lock().remove(&id);
+    }
+
+    pub fn held(&self) -> usize {
+        self.held.lock().len()
+    }
+
+    pub fn expiry(&self) -> u64 {
+        self.expiry.load(Ordering::Relaxed)
+    }
+
+    /// Changes the lock lifetime; the tests shorten it.
+    pub fn set_expiry(&self, seconds: u64) {
+        self.expiry.store(seconds.max(1), Ordering::Relaxed);
+    }
+}
+
+impl Default for TaskLocks {
+    fn default() -> Self {
+        Self {
+            held: Default::default(),
+            stopping: AtomicBool::new(false),
+            expiry: std::sync::atomic::AtomicU64::new(Self::DEFAULT_EXPIRY),
+        }
+    }
+}

crates/common/src/lib.rs

@@ -166,6 +166,10 @@ pub struct Data {
     pub logos: Mutex<AHashMap<Box<str>, LogoCache>>,
 
     pub smtp_connectors: TlsConnectors,
+
+    // inbuxa: the objects that failed to build when the running settings
+    // were built, at boot or by the last applied reload (see reload_registry)
+    pub build_errors: Mutex<AHashSet<registry::types::id::ObjectId>>,
 }
 
 #[derive(Clone)]
@@ -279,6 +283,8 @@ pub struct HttpAuthCache {
 pub struct Ipc {
     pub push_tx: mpsc::Sender<PushEvent>,
     pub task_tx: Arc<Notify>,
+    // inbuxa: task locks held by this node, released on a graceful stop
+    pub task_locks: Arc<crate::ipc::TaskLocks>,
     pub queue_tx: mpsc::Sender<QueueEvent>,
     pub report_tx: mpsc::Sender<ReportingEvent>,
     pub broadcast_tx: Option<mpsc::Sender<BroadcastEvent>>,

crates/common/src/manager/boot.rs

@@ -240,6 +240,9 @@ impl BootManager {
                     .parse_tcp_acceptors(&mut bootstrap, inner.clone())
                     .await;
 
+                // inbuxa: a reload isn't refused over objects that failed here
+                inner.build_server().record_build_errors(&bootstrap.errors);
+
                 BootManager {
                     inner,
                     bootstrap,
@@ -297,6 +300,7 @@ pub fn build_ipc(has_pubsub: bool) -> (Ipc, IpcReceivers) {
             report_tx,
             broadcast_tx: has_pubsub.then_some(broadcast_tx),
             task_tx: Arc::new(Notify::new()),
+            task_locks: Arc::new(crate::ipc::TaskLocks::default()),
             train_task_controller: Arc::new(TrainTaskController::default()),
         },
         IpcReceivers {

crates/coordinator/Cargo.toml

@@ -8,7 +8,7 @@ store = { path =  "../store" }
 registry = { path =  "../registry" }
 trc = { path = "../trc" }
 futures = { version = "0.3", optional = true }
-tokio = { version = "1.53", features = ["sync", "fs", "io-util"] }
+tokio = { version = "1.53", features = ["sync", "fs", "io-util", "rt", "time"] }
 async-nats = { version = "0.50", default-features = false, features = ["server_2_10", "server_2_11", "aws-lc-rs"], optional = true }
 zenoh = { version = "1.10.0", default-features = false, features = ["auth_pubkey", "transport_multilink", "transport_compression", "transport_quic", "transport_tcp", "transport_tls", "transport_udp"], optional = true }
 rdkafka = { version = "0.39", features = ["cmake-build"], optional = true }

crates/coordinator/src/backend/nats/mod.rs

@@ -2,13 +2,22 @@
  * SPDX-FileCopyrightText: 2020 Stalwart Labs LLC <[email protected]>
  *
  * SPDX-License-Identifier: AGPL-3.0-only OR LicenseRef-SEL
+ *
+ * Modified by Coffey Labs in 2026 for INBUXA.
  */
 
-use std::sync::Arc;
+use std::{
+    sync::{
+        Arc,
+        atomic::{AtomicBool, Ordering},
+    },
+    time::Duration,
+};
 
 use crate::Coordinator;
 use async_nats::Client;
 use registry::schema::structs::NatsCoordinator;
+use trc::ClusterEvent;
 
 pub mod pubsub;
 
@@ -47,9 +56,116 @@ impl NatsPubSub {
             opts = opts.token(credentials);
         }
 
+        // inbuxa: connect in the background and keep trying, so a node that
+        // starts while NATS is down still joins the cluster once NATS is
+        // back, instead of running without a coordinator until restarted;
+        // and report the connection going and coming back
+        let reporter = Arc::new(Reporter::default());
+        opts = opts.retry_on_initial_connect().event_callback({
+            let reporter = reporter.clone();
+            move |event| {
+                let reporter = reporter.clone();
+                async move { reporter.report(event) }
+            }
+        });
+        let connection_timeout = config.timeout_connection.into_inner();
+
         async_nats::connect_with_options(config.addresses.into_inner(), opts)
             .await
-            .map(|client| Coordinator::Nats(Arc::new(NatsPubSub { client })))
+            .map(|client| {
+                reporter.watch_first_connection(client.clone(), connection_timeout);
+                Coordinator::Nats(Arc::new(NatsPubSub { client }))
+            })
             .map_err(|err| format!("Failed to connect to Nats: {}", err))
     }
+
+    /// inbuxa: whether the client is connected to a NATS server right now.
+    pub fn is_connected(&self) -> bool {
+        matches!(
+            self.client.connection_state(),
+            async_nats::connection::State::Connected
+        )
+    }
+}
+
+/// inbuxa: reports the client's connection events as the server's own.
+#[derive(Default)]
+struct Reporter {
+    connected_once: AtomicBool,
+    // A failed attempt raises an error each time the client retries, every
+    // few seconds while NATS is down: report the first after each change
+    error_reported: AtomicBool,
+}
+
+impl Reporter {
+    fn report(&self, event: async_nats::Event) {
+        match event {
+            async_nats::Event::Connected => {
+                self.connected_once.store(true, Ordering::Relaxed);
+                self.error_reported.store(false, Ordering::Relaxed);
+                trc::event!(Cluster(ClusterEvent::CoordinatorConnected), Type = "nats");
+            }
+            async_nats::Event::Disconnected => {
+                self.error_reported.store(false, Ordering::Relaxed);
+                trc::event!(
+                    Cluster(ClusterEvent::CoordinatorDisconnected),
+                    Type = "nats",
+                    Details = "Connection lost; reconnecting in the background",
+                );
+            }
+            async_nats::Event::Closed => {
+                trc::event!(
+                    Cluster(ClusterEvent::CoordinatorDisconnected),
+                    Type = "nats",
+                    Details = "Connection closed; no further attempts will be made",
+                );
+            }
+            async_nats::Event::ClientError(async_nats::ClientError::MaxReconnects) => {
+                trc::event!(
+                    Cluster(ClusterEvent::CoordinatorDisconnected),
+                    Type = "nats",
+                    Details = "Gave up reconnecting (maxReconnects reached)",
+                );
+            }
+            async_nats::Event::ClientError(err) => {
+                if !self.error_reported.swap(true, Ordering::Relaxed) {
+                    trc::event!(
+                        Cluster(ClusterEvent::CoordinatorError),
+                        Type = "nats",
+                        Details = "Connection attempt failed; retrying",
+                        Reason = err.to_string(),
+                    );
+                }
+            }
+            event => {
+                trc::event!(
+                    Cluster(ClusterEvent::CoordinatorError),
+                    Type = "nats",
+                    Details = event.to_string(),
+                );
+            }
+        }
+    }
+
+    /// The first connection is made in the background, so say so when it
+    /// hasn't been made within the connection timeout. The client keeps
+    /// trying, and reports the connection when it comes.
+    fn watch_first_connection(self: &Arc<Self>, client: Client, timeout: Duration) {
+        let reporter = self.clone();
+        tokio::spawn(async move {
+            tokio::time::sleep(timeout).await;
+            if !reporter.connected_once.load(Ordering::Relaxed)
+                && !matches!(
+                    client.connection_state(),
+                    async_nats::connection::State::Connected
+                )
+            {
+                trc::event!(
+                    Cluster(ClusterEvent::CoordinatorDisconnected),
+                    Type = "nats",
+                    Details = "Not connected at startup; retrying in the background",
+                );
+            }
+        });
+    }
 }

crates/coordinator/src/dispatch.rs

@@ -2,6 +2,8 @@
  * SPDX-FileCopyrightText: 2020 Stalwart Labs LLC <[email protected]>
  *
  * SPDX-License-Identifier: AGPL-3.0-only OR LicenseRef-SEL
+ *
+ * Modified by Coffey Labs in 2026 for INBUXA.
  */
 
 use crate::{Coordinator, Msg, PubSubStream};
@@ -43,6 +45,17 @@ impl Coordinator {
     pub fn is_none(&self) -> bool {
         matches!(self, Coordinator::None)
     }
+
+    /// inbuxa: whether the coordinator is connected right now, for the
+    /// backends that track it (NATS); `None` for the others and when no
+    /// coordinator is configured.
+    pub fn is_connected(&self) -> Option<bool> {
+        match self {
+            #[cfg(feature = "nats")]
+            Coordinator::Nats(store) => Some(store.is_connected()),
+            _ => None,
+        }
+    }
 }
 
 impl PubSubStream {

crates/http/src/request.rs

@@ -562,6 +562,27 @@ impl ParseHttp for Server {
                         })
                         .into_http_response());
                     }
+                    // inbuxa: the cluster coordinator's connection, for
+                    // monitoring. It stays out of live and ready on purpose:
+                    // a node without its coordinator still serves mail, and
+                    // failing those would have an orchestrator restart, or
+                    // take out of service, every node at once when the
+                    // coordinator goes down
+                    "cluster" => {
+                        let coordinator = &self.core.storage.coordinator;
+                        let (status, state) = match coordinator.is_connected() {
+                            Some(true) => (StatusCode::OK, "connected"),
+                            Some(false) => (StatusCode::SERVICE_UNAVAILABLE, "disconnected"),
+                            None if coordinator.is_none() => (StatusCode::OK, "none"),
+                            None => (StatusCode::OK, "unknown"),
+                        };
+                        return Ok(http_proto::JsonResponse::with_status(
+                            status,
+                            serde_json::json!({ "coordinator": state }),
+                        )
+                        .no_cache()
+                        .into_http_response());
+                    }
                     _ => (),
                 }
             }

crates/jmap/src/registry/mapping/action.rs

@@ -2,6 +2,8 @@
  * SPDX-FileCopyrightText: 2020 Stalwart Labs LLC <[email protected]>
  *
  * SPDX-License-Identifier: AGPL-3.0-only OR LicenseRef-SEL
+ *
+ * Modified by Coffey Labs in 2026 for INBUXA.
  */
 
 use crate::registry::mapping::{RegistrySetResponse, map_bootstrap_error};
@@ -99,7 +101,7 @@ pub(crate) async fn action_set(
                 } else {
                     set.response
                         .not_created
-                        .append(id, map_bootstrap_error(result.errors));
+                        .append(id, reload_refused(result.errors));
                 }
             }
             Action::InvalidateCaches => {
@@ -573,3 +575,34 @@ async fn dmarc_troubleshoot(
 
     Some(request)
 }
+
+/// inbuxa: a refused reload names the object that stopped it and says the
+/// settings weren't applied; upstream passed on the first error's bare message
+/// ("Invalid address: ..."), which read like a problem with the request.
+fn reload_refused(errors: Vec<registry::types::error::Error>) -> SetError<Property> {
+    use registry::types::error::Error;
+    let more = errors.len().saturating_sub(1);
+    let mut description = match errors.first() {
+        Some(Error::Build { object_id, message }) => format!("{object_id}: {message}"),
+        Some(Error::Validation { object_id, errors }) => format!(
+            "{object_id}: {}",
+            errors
+                .iter()
+                .map(|err| err.to_string())
+                .collect::<Vec<_>>()
+                .join("; ")
+        ),
+        Some(Error::Internal {
+            object_id: Some(object_id),
+            error,
+        }) => format!("{object_id}: {error}"),
+        Some(Error::Internal { error, .. }) => error.to_string(),
+        Some(Error::NotFound { object_id }) => format!("{object_id} was not found"),
+        None => String::new(),
+    };
+    description.insert_str(0, "Settings were not reloaded. ");
+    if more > 0 {
+        description.push_str(&format!(" ({more} more in the server log.)"));
+    }
+    map_bootstrap_error(errors).with_description(description)
+}

crates/jmap/src/registry/set.rs

@@ -947,10 +947,14 @@ impl RegistrySet for Server {
                     self.cluster_broadcast(common::ipc::BroadcastEvent::RegistryChange(change))
                         .await;
                 }
-                Ok(_) => trc::event!(
+                Ok(reload) => {
+                    // inbuxa: name what stopped it
+                    reload.log();
+                    trc::event!(
                         Registry(trc::RegistryEvent::BuildWarning),
                         Details = "Settings didn't reload after a directory change",
-                ),
+                    )
+                }
                 Err(err) => {
                     trc::error!(err.details("Failed to reload directories"));
                 }

crates/main/src/main.rs

@@ -109,6 +109,10 @@ async fn main() -> std::io::Result<()> {
     // Wait for shutdown signal
     wait_for_shutdown().await;
 
+    // inbuxa: hand back the task locks this node holds, so other nodes can
+    // run those tasks now rather than when the locks expire
+    services::task_manager::lock::release_task_locks(&inner.build_server()).await;
+
     // Shutdown collector
     Collector::shutdown();
 

crates/services/src/task_manager/index.rs

@@ -91,7 +91,15 @@ impl SearchIndexTask for Server {
                             build_contact_document(self, account_id, document_id).await
                         }
                         IndexDocumentType::File => {
-                            // File indexing not implemented yet
+                            // File indexing not implemented yet. inbuxa: still
+                            // one result per task: update_tasks pairs them by
+                            // position, and a missing one shifts every result
+                            // after it onto the wrong task
+                            results.push(IndexTaskResult {
+                                task_type: TaskType::Insert,
+                                index: task.document_type,
+                                result: TaskResult::Ignored,
+                            });
                             continue;
                         }
                     };

crates/services/src/task_manager/lock.rs

@@ -2,6 +2,8 @@
  * SPDX-FileCopyrightText: 2020 Stalwart Labs LLC <[email protected]>
  *
  * SPDX-License-Identifier: AGPL-3.0-only OR LicenseRef-SEL
+ *
+ * Modified by Coffey Labs in 2026 for INBUXA.
  */
 
 use crate::task_manager::*;
@@ -13,13 +15,21 @@ pub trait TaskLockManager: Sync + Send {
 
 impl TaskLockManager for Server {
     async fn try_lock_task(&self, id: u64) -> bool {
+        // inbuxa: a node that is stopping claims nothing new
+        let locks = &self.inner.ipc.task_locks;
+        if locks.is_stopping() {
+            return false;
+        }
+
         match self
             .in_memory_store()
-            .try_lock(KV_LOCK_TASK, &id.to_be_bytes(), DEFAULT_LOCK_EXPIRY)
+            .try_lock(KV_LOCK_TASK, &id.to_be_bytes(), locks.expiry())
             .await
         {
             Ok(result) => {
-                if !result {
+                if result {
+                    locks.insert(id);
+                } else {
                     trc::event!(
                         TaskManager(TaskManagerEvent::TaskLocked),
                         Id = id,
@@ -48,5 +58,27 @@ impl TaskLockManager for Server {
                     .caused_by(trc::location!())
             );
         }
+        self.inner.ipc.task_locks.remove(id);
     }
 }
+
+/// inbuxa: on a graceful stop, stops claiming tasks and releases every task
+/// lock this node holds, so the rest of the cluster can pick the tasks up at
+/// once instead of after the lock expires. Returns how many were released.
+pub async fn release_task_locks(server: &Server) -> usize {
+    let ids = server.inner.ipc.task_locks.stop();
+    for id in &ids {
+        if let Err(err) = server
+            .in_memory_store()
+            .remove_lock(KV_LOCK_TASK, &id.to_be_bytes())
+            .await
+        {
+            trc::error!(
+                err.details("Failed to release task lock on shutdown")
+                    .ctx(trc::Key::Id, *id)
+                    .caused_by(trc::location!())
+            );
+        }
+    }
+    ids.len()
+}

crates/services/src/task_manager/manager.rs

@@ -2,6 +2,8 @@
  * SPDX-FileCopyrightText: 2020 Stalwart Labs LLC <[email protected]>
  *
  * SPDX-License-Identifier: AGPL-3.0-only OR LicenseRef-SEL
+ *
+ * Modified by Coffey Labs in 2026 for INBUXA.
  */
 
 use crate::task_manager::acme::AcmeTask;
@@ -18,7 +20,7 @@ use crate::task_manager::report::{self, SubmitReportTask};
 use crate::task_manager::restore_item::RestoreItemTask;
 use crate::task_manager::spam_classifier::SpamFilterMaintenanceTask;
 use crate::task_manager::{
-    DEFAULT_LOCK_EXPIRY, Locked, QUEUE_REFRESH_INTERVAL, TaskDetails, TaskFailureType, TaskInfo,
+    CLAIM_RECHECK_INTERVAL, Locked, QUEUE_REFRESH_INTERVAL, TaskDetails, TaskFailureType, TaskInfo,
     TaskJob, TaskManagerIpc, TaskResult,
 };
 use common::BuildServer;
@@ -124,72 +126,47 @@ pub fn spawn_task_manager(inner: Arc<Inner>) {
                     let server = inner.build_server();
                     let batch_size = server.core.email.index_batch_size;
                     let mut batch = Vec::with_capacity(batch_size);
-                    match server
+                    if let Some(task) = fetch_task(&server, job).await {
-                        .store()
+                        batch.push(task);
-                        .get_value::<Task>(ValueKey::from(ValueClass::TaskQueue(
-                            TaskQueueClass::Task { id: job.id },
-                        )))
-                        .await
-                    {
-                        Ok(Some(task)) => {
-                            batch.push(TaskDetails { task, info: job });
-                        }
-                        Ok(None) => {
-                            trc::event!(
-                                TaskManager(TaskManagerEvent::TaskIgnored),
-                                Id = job.id,
-                                Reason = "Task not found in store, likely already processed.",
-                            );
-                        }
-                        Err(err) => {
-                            trc::error!(
-                                err.id(job.id)
-                                    .details("Failed to retrieve task details.")
-                                    .caused_by(trc::location!())
-                            );
-                        }
                     }
 
                     while batch.len() < batch_size {
                         match rx.try_recv() {
                             Ok(job) => {
-                                match server
+                                if let Some(task) = fetch_task(&server, job).await {
-                                    .store()
+                                    batch.push(task);
-                                    .get_value::<Task>(ValueKey::from(ValueClass::TaskQueue(
-                                        TaskQueueClass::Task { id: job.id },
-                                    )))
-                                    .await
-                                {
-                                    Ok(Some(task)) => {
-                                        batch.push(TaskDetails { task, info: job });
-                                    }
-                                    Ok(None) => {
-                                        trc::event!(
-                                            TaskManager(TaskManagerEvent::TaskIgnored),
-                                            Id = job.id,
-                                            Reason = "Task not found in store, likely already processed.",
-                                        );
-                                    }
-                                    Err(err) => {
-                                        trc::error!(
-                                            err.id(job.id)
-                                                .details("Failed to retrieve task details.")
-                                                .caused_by(trc::location!())
-                                        );
-                                    }
                                 }
                             }
                             Err(_) => break,
                         }
                     }
 
-                    // Dispatch
+                    // Dispatch. inbuxa: on a task of its own, so a panic
+                    // releases the batch's locks and leaves this worker
+                    // running; a dead worker would keep claiming tasks it
+                    // can never run
                     let mut refresh_queue = false;
-                    let results = server.index(&batch).await.into_iter().map(|r| {
+                    let ids = batch.iter().map(|task| task.info.id).collect::<Vec<_>>();
+                    let run = {
+                        let server = server.clone();
+                        tokio::spawn(async move {
+                            let results = server.index(&batch).await;
+                            (batch, results)
+                        })
+                    };
+                    match run.await {
+                        Ok((mut batch, results)) => {
+                            let results = results.into_iter().map(|r| {
                                 refresh_queue |= r.result.is_retry();
                                 r.result
                             });
                             update_tasks(&server, &mut batch, results).await;
+                        }
+                        Err(err) => {
+                            worker_failed(&server, &ids, err).await;
+                            refresh_queue = true;
+                        }
+                    }
 
                     if refresh_queue || rx.is_empty() {
                         server.notify_task_queue();
@@ -203,83 +180,31 @@ pub fn spawn_task_manager(inner: Arc<Inner>) {
                     let server = inner.build_server();
                     let mut refresh_queue = false;
 
-                    match server
+                    if let Some(TaskDetails { task, info }) = fetch_task(&server, job).await {
-                        .store()
+                        // inbuxa: on a task of its own, as above
-                        .get_value::<Task>(ValueKey::from(ValueClass::TaskQueue(
+                        let run = {
-                            TaskQueueClass::Task { id: job.id },
+                            let server = server.clone();
-                        )))
+                            let server_instance = server_instance.clone();
-                        .await
+                            tokio::spawn(async move {
-                    {
+                                let result = run_task(&server, &task, server_instance).await;
-                        Ok(Some(task)) => {
+                                (task, result)
-                            let result = match &task {
+                            })
-                                Task::CalendarAlarmEmail(task) => {
-                                    server.send_email_alarm(task, server_instance.clone()).await
-                                }
-                                Task::CalendarAlarmNotification(task) => {
-                                    server.send_display_alarm(task).await
-                                }
-                                Task::CalendarItipMessage(task) => {
-                                    server.send_imip(task, server_instance.clone()).await
-                                }
-                                Task::MergeThreads(task) => server.merge_threads(task).await,
-                                Task::DmarcReport(task) => {
-                                    server
-                                        .submit_report(report::ReportId::Dmarc(task.report_id.id()))
-                                        .await
-                                }
-                                Task::TlsReport(task) => {
-                                    server
-                                        .submit_report(report::ReportId::Tls(task.report_id.id()))
-                                        .await
-                                }
-                                Task::RestoreArchivedItem(task) => server.restore_item(task).await,
-                                Task::DestroyAccount(task) => server.destroy_account(task).await,
-                                Task::AccountMaintenance(task) => {
-                                    server.account_maintenance(task).await
-                                }
-                                Task::TenantMaintenance(task) => {
-                                    server.tenant_maintenance(task).await
-                                }
-                                Task::StoreMaintenance(task) => {
-                                    server.store_maintenance(task).await
-                                }
-                                Task::SpamFilterMaintenance(task) => {
-                                    Box::pin(server.spam_filter_maintenance(task)).await
-                                }
-                                Task::AcmeRenewal(task) => server.acme_management(task).await,
-                                Task::DkimManagement(task_dkim_rotation) => {
-                                    server.dkim_management(task_dkim_rotation).await
-                                }
-                                Task::DnsManagement(task_dns_management) => {
-                                    server.dns_management(task_dns_management).await
-                                }
-                                Task::IndexDocument(_)
-                                | Task::UnindexDocument(_)
-                                | Task::IndexTrace(_) => unreachable!(),
                         };
-
+                        match run.await {
+                            Ok((task, result)) => {
                                 refresh_queue = result.is_retry();
 
                                 update_tasks(
                                     &server,
-                                &mut [TaskDetails { task, info: job }],
+                                    &mut [TaskDetails { task, info }],
                                     vec![result],
                                 )
                                 .await;
                             }
-                        Ok(None) => {
-                            trc::event!(
-                                TaskManager(TaskManagerEvent::TaskIgnored),
-                                Id = job.id,
-                                Reason = "Task not found in store, likely already processed.",
-                            );
-                        }
                             Err(err) => {
-                            trc::error!(
+                                worker_failed(&server, &[info.id], err).await;
-                                err.id(job.id)
+                                refresh_queue = true;
-                                    .details("Failed to retrieve task details.")
+                            }
-                                    .caused_by(trc::location!())
-                            );
                         }
                     }
 
@@ -318,6 +243,13 @@ pub(crate) trait TaskQueueManager: Sync + Send {
 
 impl TaskQueueManager for Server {
     async fn process_tasks(&self, ipc: &mut TaskManagerIpc) -> Duration {
+        // inbuxa: a node that is stopping has released its locks and claims
+        // nothing new
+        let task_locks = &self.inner.ipc.task_locks;
+        if task_locks.is_stopping() {
+            return Duration::from_secs(QUEUE_REFRESH_INTERVAL);
+        }
+        let lock_expiry = task_locks.expiry();
         let now_timestamp = now();
         let from_key = ValueKey::<ValueClass> {
             account_id: 0,
@@ -393,9 +325,7 @@ impl TaskQueueManager for Server {
                                     let locked = entry.get_mut();
                                     if locked.expires <= now || locked.due < task_due {
                                         locked.expires = Instant::now()
-                                            + std::time::Duration::from_secs(
+                                            + std::time::Duration::from_secs(lock_expiry + 1);
-                                                DEFAULT_LOCK_EXPIRY + 1,
-                                            );
                                         locked.due = task_due;
                                         tasks.push((
                                             TaskJob {
@@ -411,9 +341,7 @@ impl TaskQueueManager for Server {
                                 Entry::Vacant(entry) => {
                                     entry.insert(Locked {
                                         expires: Instant::now()
-                                            + std::time::Duration::from_secs(
+                                            + std::time::Duration::from_secs(lock_expiry + 1),
-                                                DEFAULT_LOCK_EXPIRY + 1,
-                                            ),
                                         due: task_due,
                                         revision: ipc.revision,
                                     });
@@ -464,12 +392,26 @@ impl TaskQueueManager for Server {
             let tx = &ipc.txs[task_type_idx as usize];
 
             if tx.capacity() > 0 {
-                if self.try_lock_task(task_job.id).await && tx.send(task_job).await.is_err() {
+                let id = task_job.id;
+                if !self.try_lock_task(id).await {
+                    // inbuxa: another node holds the task. Look again after a
+                    // short while rather than a full lock lifetime from now:
+                    // the holder may have claimed it after this scan began,
+                    // or run on a clock ahead of this one, and waiting the
+                    // whole lifetime again would leave the task stuck for
+                    // another hour past its lock if that holder died
+                    if let Some(locked) = ipc.locked.get_mut(&id) {
+                        locked.expires =
+                            Instant::now() + Duration::from_secs(claim_recheck_interval(lock_expiry));
+                    }
+                } else if tx.send(task_job).await.is_err() {
                     trc::event!(
                         Server(trc::ServerEvent::ThreadError),
                         Details = "Error sending task.",
                         CausedBy = trc::location!()
                     );
+                    // inbuxa: nothing will run it here, so don't hold it
+                    self.remove_index_lock(id).await;
                 }
             } else {
                 // If the channel is full, release the lock so it can be picked up in the next iteration
@@ -481,9 +423,117 @@ impl TaskQueueManager for Server {
         let now = Instant::now();
         ipc.locked
             .retain(|_, locked| locked.expires > now && locked.revision == ipc.revision);
-        Duration::from_secs(next_event.map_or(QUEUE_REFRESH_INTERVAL, |timestamp| {
+        let sleep_for = Duration::from_secs(next_event.map_or(QUEUE_REFRESH_INTERVAL, |timestamp| {
             timestamp.saturating_sub(store::write::now())
-        }))
+        }));
+
+        // inbuxa: wake up when a claim held elsewhere is due to be tried
+        // again, rather than only on the next task or refresh
+        ipc.locked
+            .values()
+            .map(|locked| locked.expires.saturating_duration_since(now))
+            .min()
+            .map_or(sleep_for, |recheck| sleep_for.min(recheck.max(Duration::from_secs(1))))
+    }
+}
+
+async fn run_task(
+    server: &Server,
+    task: &Task,
+    server_instance: Arc<ServerInstance>,
+) -> TaskResult {
+    match task {
+        Task::CalendarAlarmEmail(task) => {
+            server.send_email_alarm(task, server_instance.clone()).await
+        }
+        Task::CalendarAlarmNotification(task) => {
+            server.send_display_alarm(task).await
+        }
+        Task::CalendarItipMessage(task) => {
+            server.send_imip(task, server_instance.clone()).await
+        }
+        Task::MergeThreads(task) => server.merge_threads(task).await,
+        Task::DmarcReport(task) => {
+            server
+                .submit_report(report::ReportId::Dmarc(task.report_id.id()))
+                .await
+        }
+        Task::TlsReport(task) => {
+            server
+                .submit_report(report::ReportId::Tls(task.report_id.id()))
+                .await
+        }
+        Task::RestoreArchivedItem(task) => server.restore_item(task).await,
+        Task::DestroyAccount(task) => server.destroy_account(task).await,
+        Task::AccountMaintenance(task) => {
+            server.account_maintenance(task).await
+        }
+        Task::TenantMaintenance(task) => {
+            server.tenant_maintenance(task).await
+        }
+        Task::StoreMaintenance(task) => {
+            server.store_maintenance(task).await
+        }
+        Task::SpamFilterMaintenance(task) => {
+            Box::pin(server.spam_filter_maintenance(task)).await
+        }
+        Task::AcmeRenewal(task) => server.acme_management(task).await,
+        Task::DkimManagement(task_dkim_rotation) => {
+            server.dkim_management(task_dkim_rotation).await
+        }
+        Task::DnsManagement(task_dns_management) => {
+            server.dns_management(task_dns_management).await
+        }
+        Task::IndexDocument(_)
+        | Task::UnindexDocument(_)
+        | Task::IndexTrace(_) => unreachable!(),
+    }
+}
+
+/// Reads a claimed task. When it is gone or can't be read, the claim is
+/// released: inbuxa: holding it would block the task, everywhere, until
+/// the lock expired.
+async fn fetch_task(server: &Server, job: TaskJob) -> Option<TaskDetails> {
+    match server
+        .store()
+        .get_value::<Task>(ValueKey::from(ValueClass::TaskQueue(TaskQueueClass::Task {
+            id: job.id,
+        })))
+        .await
+    {
+        Ok(Some(task)) => Some(TaskDetails { task, info: job }),
+        Ok(None) => {
+            trc::event!(
+                TaskManager(TaskManagerEvent::TaskIgnored),
+                Id = job.id,
+                Reason = "Task not found in store, likely already processed.",
+            );
+            server.remove_index_lock(job.id).await;
+            None
+        }
+        Err(err) => {
+            trc::error!(
+                err.id(job.id)
+                    .details("Failed to retrieve task details.")
+                    .caused_by(trc::location!())
+            );
+            server.remove_index_lock(job.id).await;
+            None
+        }
+    }
+}
+
+/// inbuxa: a task panicked: its locks are released so it runs again, here or
+/// on another node, and the worker carries on.
+async fn worker_failed(server: &Server, ids: &[u64], err: tokio::task::JoinError) {
+    trc::event!(
+        Server(trc::ServerEvent::ThreadError),
+        Details = "Task worker failed",
+        Reason = err.to_string(),
+        CausedBy = trc::location!()
+    );
+    for id in ids {
+        server.remove_index_lock(*id).await;
     }
 }
 
@@ -614,6 +664,13 @@ async fn update_tasks(
     }
 }
 
+/// inbuxa: how long to wait before trying again to claim a task another node
+/// holds: a twelfth of the lock lifetime, so five minutes for the one-hour
+/// lock, never more than that and never under a second.
+pub(crate) fn claim_recheck_interval(lock_expiry: u64) -> u64 {
+    (lock_expiry / 12).clamp(1, CLAIM_RECHECK_INTERVAL)
+}
+
 pub fn perpetual_retry_time(typ: TaskType, attempt: u64) -> Option<u64> {
     matches!(
         typ,

crates/services/src/task_manager/mod.rs

@@ -35,7 +35,9 @@ pub mod scheduler;
 pub mod spam_classifier;
 
 const QUEUE_REFRESH_INTERVAL: u64 = 60 * 5; // 5 minutes
-const DEFAULT_LOCK_EXPIRY: u64 = 60 * 60; // 1 hour
+// inbuxa: the lock lifetime (one hour) lives in common::ipc::TaskLocks, per
+// server, so a graceful stop can release the locks and the tests can shorten it
+const CLAIM_RECHECK_INTERVAL: u64 = 60 * 5; // 5 minutes
 
 pub(crate) struct TaskManagerIpc {
     txs: [mpsc::Sender<TaskJob>; TaskType::COUNT],

crates/smtp/src/inbound/milter/client.rs

@@ -2,6 +2,8 @@
  * SPDX-FileCopyrightText: 2020 Stalwart Labs LLC <[email protected]>
  *
  * SPDX-License-Identifier: AGPL-3.0-only OR LicenseRef-SEL
+ *
+ * Modified by Coffey Labs in 2026 for INBUXA.
  */
 
 use common::config::smtp::session::Milter;
@@ -25,7 +27,19 @@ impl MilterClient<TcpStream> {
     pub async fn connect(config: &Milter, session_id: u64) -> Result<Self> {
         tokio::time::timeout(config.timeout_command, async {
             let mut last_err = Error::Disconnected;
-            for addr in &config.addrs {
+            // inbuxa: a hostname is resolved here, per connection, rather
+            // than while the settings are built
+            let resolved;
+            let addrs = if config.addrs.is_empty() {
+                resolved = tokio::net::lookup_host((config.hostname.as_str(), config.port))
+                    .await
+                    .map_err(Error::Io)?
+                    .collect::<Vec<_>>();
+                &resolved
+            } else {
+                &config.addrs
+            };
+            for addr in addrs {
                 match TcpStream::connect(addr).await {
                     Ok(stream) => {
                         return Ok(MilterClient {

crates/spam-filter/src/modules/pyzor.rs

@@ -48,16 +48,21 @@ pub(crate) async fn pyzor_check(
     // Send message to address. inbuxa: in tests, a fixed table answers
     // instead of a public server (test_response).
     #[cfg(not(feature = "test_mode"))]
-    let response = pyzor_send_message(config.address, config.timeout, &request).await;
+    let response = match tokio::time::timeout(config.timeout, config.address()).await {
+        Ok(Ok(address)) => pyzor_send_message(address, config.timeout, &request).await,
+        Ok(Err(err)) => Err(err),
+        Err(_) => Err(std::io::Error::new(
+            std::io::ErrorKind::TimedOut,
+            "Timed out resolving the Pyzor server",
+        )),
+    };
     #[cfg(feature = "test_mode")]
     let response = std::io::Result::Ok(test_response(&request));
 
-    response
+    response.map(Into::into).map_err(|err| {
-        .map(Into::into)
-        .map_err(|err| {
         trc::SpamEvent::PyzorError
             .into_err()
-                .ctx(trc::Key::Url, config.address.to_string())
+            .ctx(trc::Key::Url, format!("{}:{}", config.host, config.port))
             .reason(err)
             .details("Pyzor failed")
     })

crates/store/src/backend/mysql/search.rs

@@ -2,6 +2,8 @@
  * SPDX-FileCopyrightText: 2020 Stalwart Labs LLC <[email protected]>
  *
  * SPDX-License-Identifier: AGPL-3.0-only OR LicenseRef-SEL
+ *
+ * Modified by Coffey Labs in 2026 for INBUXA.
  */
 
 use crate::{
@@ -19,7 +21,7 @@ use crate::{
     write::SearchIndex,
 };
 use mysql_async::{IsolationLevel, TxOpts, Value, prelude::Queryable};
-use nlp::tokenizers::word::WordTokenizer;
+use nlp::{language::Language, tokenizers::word::WordTokenizer};
 use std::fmt::Write;
 
 impl MysqlStore {
@@ -146,6 +148,20 @@ impl MysqlStore {
     }
 }
 
+// inbuxa: InnoDB's default full-text stopword list
+// (INFORMATION_SCHEMA.INNODB_FT_DEFAULT_STOPWORD) and innodb_ft_min_token_size
+// default; words outside these are not in a FULLTEXT index.
+const FT_STOPWORDS: &[&str] = &[
+    "a", "about", "an", "are", "as", "at", "be", "by", "com", "de", "en", "for", "from", "how",
+    "i", "in", "is", "it", "la", "of", "on", "or", "that", "the", "this", "to", "was", "what",
+    "when", "where", "who", "will", "with", "und", "www",
+];
+const FT_MIN_TOKEN_SIZE: usize = 3;
+
+fn is_ft_indexed(word: &str) -> bool {
+    word.chars().count() >= FT_MIN_TOKEN_SIZE && !FT_STOPWORDS.contains(&word)
+}
+
 fn build_filter(query: &mut String, filters: &[SearchFilter]) -> Vec<Value> {
     if filters.is_empty() {
         return Vec::new();
@@ -171,30 +187,77 @@ fn build_filter(query: &mut String, filters: &[SearchFilter]) -> Vec<Value> {
 
                 if field.is_text() && matches!(op, SearchOperator::Equal | SearchOperator::Contains)
                 {
-                    let (value, mode) = match (value, op) {
+                    let (value, mode, unindexed) = match (value, op) {
-                        (SearchValue::Text { value, .. }, SearchOperator::Equal) => {
+                        (SearchValue::Text { value, .. }, SearchOperator::Equal) => (
-                            (Value::Bytes(format!("{value:?}").into_bytes()), "BOOLEAN")
+                            Value::Bytes(format!("{value:?}").into_bytes()),
-                        }
+                            "BOOLEAN",
-                        (SearchValue::Text { value, .. }, ..) => {
+                            Vec::new(),
+                        ),
+                        (SearchValue::Text { value, language }, ..) => {
                             let mut text_query = String::with_capacity(value.len() + 1);
+                            let mut unindexed = Vec::new();
 
                             for item in WordTokenizer::new(value, MAX_TOKEN_LENGTH) {
+                                // inbuxa: InnoDB never indexes stopwords ("com",
+                                // "de", "www", ...) or words under
+                                // innodb_ft_min_token_size, and a required
+                                // (+word) term it has not indexed matches no row,
+                                // so "example.com" or "[email protected]" found
+                                // nothing. Such words are matched with a
+                                // word-boundary REGEXP instead.
+                                if is_ft_indexed(&item.word) {
                                     if !text_query.is_empty() {
                                         text_query.push(' ');
                                     }
                                     text_query.push('+');
                                     text_query.push_str(&item.word);
+                                } else {
+                                    unindexed.push(item.word);
+                                }
                             }
 
-                            (Value::Bytes(text_query.into_bytes()), "BOOLEAN")
+                            // For language text (bodies, subjects) the unindexed
+                            // words are noise words and only checked when nothing
+                            // else is left to match; keyword text (addresses,
+                            // contact fields) checks every word, as the other
+                            // backends do.
+                            if !text_query.is_empty() && !matches!(language, Language::None) {
+                                unindexed.clear();
+                            }
+
+                            (Value::Bytes(text_query.into_bytes()), "BOOLEAN", unindexed)
                         }
                         _ => {
                             debug_assert!(false, "Invalid search value for text field");
                             continue;
                         }
                     };
-                    let _ = write!(query, "MATCH({}) AGAINST(? IN {mode} MODE)", field.column());
+                    if unindexed.is_empty() {
+                        let _ =
+                            write!(query, "MATCH({}) AGAINST(? IN {mode} MODE)", field.column());
                         values.push(value);
+                    } else {
+                        query.push('(');
+                        let is_empty = matches!(&value, Value::Bytes(v) if v.is_empty());
+                        if !is_empty {
+                            let _ = write!(
+                                query,
+                                "MATCH({}) AGAINST(? IN {mode} MODE) AND ",
+                                field.column()
+                            );
+                            values.push(value);
+                        }
+                        for (i, word) in unindexed.iter().enumerate() {
+                            if i > 0 {
+                                query.push_str(" AND ");
+                            }
+                            let _ = write!(query, "{} REGEXP ?", field.column());
+                            values.push(Value::Bytes(
+                                format!("(^|[^[:alnum:]]){word}([^[:alnum:]]|$)").into_bytes(),
+                            ));
+                        }
+                        query.push(')');
+                    }
                 } else if let SearchValue::KeyValues(kv) = value {
                     let (key, value) = kv.iter().next().unwrap();
 

crates/store/src/backend/postgres/search.rs

@@ -2,20 +2,25 @@
  * SPDX-FileCopyrightText: 2020 Stalwart Labs LLC <[email protected]>
  *
  * SPDX-License-Identifier: AGPL-3.0-only OR LicenseRef-SEL
+ *
+ * Modified by Coffey Labs in 2026 for INBUXA.
  */
 
 use crate::{
-    backend::postgres::{
+    backend::{
+        MAX_TOKEN_LENGTH,
+        postgres::{
             DELETE_CHUNK_SIZE, MIN_DELETE_CHUNK_SIZE, PostgresStore, PsqlSearchField, into_error,
             into_pool_error, is_timeout_error,
         },
+    },
     search::{
         IndexDocument, SearchComparator, SearchDocumentId, SearchFilter, SearchOperator,
         SearchQuery, SearchValue,
     },
     write::SearchIndex,
 };
-use nlp::language::Language;
+use nlp::{language::Language, tokenizers::space::SpaceTokenizer};
 use std::fmt::Write;
 use tokio_postgres::{
     IsolationLevel,
@@ -43,6 +48,19 @@ impl PostgresStore {
             let primary_keys = index.primary_keys();
             let all_fields = index.all_fields();
             let fields = document.fields;
+            // inbuxa: keyword text (addresses, contact fields, ...) is split into
+            // words before it reaches the text parser, see keyword_terms().
+            let keywords = primary_keys
+                .iter()
+                .chain(all_fields)
+                .map(|field| match fields.get(field) {
+                    Some(SearchValue::Text {
+                        value,
+                        language: Language::None,
+                    }) if field.is_text() => Some(keyword_terms(value)),
+                    _ => None,
+                })
+                .collect::<Vec<_>>();
             let mut values = Vec::with_capacity(fields.len() + 2);
             let mut query = format!("INSERT INTO {} (", index.psql_table());
 
@@ -74,7 +92,20 @@ impl PostgresStore {
                         (0, PG_UNSTEMMED_LANG)
                     };
 
-                    if field.is_text() {
+                    if let Some(keywords) = &keywords[i] {
+                        let _ = write!(&mut query, "to_tsvector('{language}',{value_ref})");
+                        values.push(keywords as &(dyn ToSql + Sync));
+                        if field.sort_column().is_some() {
+                            let value_ref = format!("${}", values.len() + 1);
+                            if text_len > 255 {
+                                let _ = write!(&mut query, ",left({value_ref},255)");
+                            } else {
+                                let _ = write!(&mut query, ",{value_ref}");
+                            }
+                            values.push(value as &(dyn ToSql + Sync));
+                        }
+                        continue;
+                    } else if field.is_text() {
                         let _ = write!(&mut query, "to_tsvector('{language}',{value_ref})");
                     } else if text_len > 512 {
                         query.push_str("left(");
@@ -134,6 +165,7 @@ impl PostgresStore {
     ) -> trc::Result<Vec<R>> {
         let mut query = format!("SELECT {} FROM {}", R::field().column(), index.psql_table());
         let params = self.build_filter(&mut query, filters);
+        let params = params.iter().map(SqlParam::as_sql).collect::<Vec<_>>();
         if !sort.is_empty() {
             build_sort(&mut query, sort);
         }
@@ -155,6 +187,7 @@ impl PostgresStore {
         let table = filter.index.psql_table();
         let mut where_clause = String::new();
         let params = self.build_filter(&mut where_clause, &filter.filters);
+        let params = params.iter().map(SqlParam::as_sql).collect::<Vec<_>>();
         let conn = self.conn_pool.get().await.map_err(into_pool_error)?;
         let s = conn
             .prepare_cached(&format!("DELETE FROM {table}{where_clause}"))
@@ -196,7 +229,7 @@ impl PostgresStore {
         &self,
         query: &mut String,
         filters: &'x [SearchFilter],
-    ) -> Vec<&'x (dyn ToSql + Sync)> {
+    ) -> Vec<SqlParam<'x>> {
         if filters.is_empty() {
             return Vec::new();
         }
@@ -237,6 +270,10 @@ impl PostgresStore {
 
                         if matches!(language, Language::None) {
                             let _ = write!(query, "@@ {method}('{config}', ${value_pos})");
+                            if let SearchValue::Text { value, .. } = value {
+                                values.push(SqlParam::Owned(keyword_terms(value)));
+                                continue;
+                            }
                         } else {
                             let _ = write!(query, "@@ ({method}('{config}', ${value_pos})");
                             for fallback in [PG_FALLBACK_LANG, PG_UNSTEMMED_LANG] {
@@ -247,18 +284,18 @@ impl PostgresStore {
                             }
                             query.push(')');
                         }
-                        values.push(value as &(dyn ToSql + Sync));
+                        values.push(SqlParam::Ref(value));
                     } else if let SearchValue::KeyValues(kv) = value {
                         query.push_str(field.column());
                         query.push(' ');
 
                         let (key, value) = kv.iter().next().unwrap();
-                        values.push(key as &(dyn ToSql + Sync));
+                        values.push(SqlParam::Ref(key));
 
                         if !value.is_empty() {
                             let _ = write!(query, "->> ${value_pos} ");
                             op.write_pqsql(query, values.len() + 1);
-                            values.push(value as &(dyn ToSql + Sync));
+                            values.push(SqlParam::Ref(value));
                         } else {
                             let _ = write!(query, " ? ${value_pos}");
                         }
@@ -267,7 +304,7 @@ impl PostgresStore {
                         query.push(' ');
 
                         op.write_pqsql(query, value_pos);
-                        values.push(value as &(dyn ToSql + Sync));
+                        values.push(SqlParam::Ref(value));
                     }
                 }
                 SearchFilter::And | SearchFilter::Or => {
@@ -321,6 +358,38 @@ impl PostgresStore {
     }
 }
 
+// inbuxa: PostgreSQL's text parser keeps "[email protected]" (and host names,
+// URLs, file paths, ...) as a single token, so a search for "user" or
+// "example.com" never matched an address. Keyword text is split into words the
+// same way the built-in index splits it (SpaceTokenizer: lowercase runs of
+// alphanumerics) on both the indexing and the query side, so a full address,
+// its local part, its domain and the display-name words all match, as they do
+// on the other backends.
+pub(crate) fn keyword_terms(value: &str) -> String {
+    let mut terms = String::with_capacity(value.len());
+    for token in SpaceTokenizer::new(value, MAX_TOKEN_LENGTH) {
+        if !terms.is_empty() {
+            terms.push(' ');
+        }
+        terms.push_str(&token);
+    }
+    terms
+}
+
+pub(super) enum SqlParam<'x> {
+    Ref(&'x (dyn ToSql + Sync)),
+    Owned(String),
+}
+
+impl SqlParam<'_> {
+    fn as_sql(&self) -> &(dyn ToSql + Sync) {
+        match self {
+            SqlParam::Ref(value) => *value,
+            SqlParam::Owned(value) => value,
+        }
+    }
+}
+
 fn build_sort(query: &mut String, sort: &[SearchComparator]) {
     query.push_str(" ORDER BY ");
     for (i, comparator) in sort.iter().enumerate() {

crates/trc/src/event/enums.rs

@@ -10,8 +10,9 @@
 
 // inbuxa: 637 to 641 are the fork's SCIM events (SCIM-54); 642 is
 // auth.legacy-protocol-refused (legacy-protocols LP-6); 643 is
-// security.legacy-protocols-changed (LP-8)
+// security.legacy-protocols-changed (LP-8); 644 to 646 are the cluster
-pub const TOTAL_EVENT_COUNT: usize = 644;
+// coordinator's connection events
+pub const TOTAL_EVENT_COUNT: usize = 647;
 pub const TOTAL_METRIC_COUNT: usize = 369;
 
 #[derive(Debug, Clone, Copy, PartialEq, Eq, Hash)]
@@ -150,6 +151,10 @@ pub enum ClusterEvent {
     MessageSkipped = 47,
     MessageInvalid = 49,
     NodeIdRenewed = 275,
+    // inbuxa: the coordinator's connection
+    CoordinatorConnected = 644,
+    CoordinatorDisconnected = 645,
+    CoordinatorError = 646,
 }
 
 #[derive(Debug, Clone, Copy, PartialEq, Eq, Hash)]

crates/trc/src/event/enums_impl.rs

@@ -81,6 +81,10 @@ impl EventType {
             b"cluster.message-skipped" => EventType::Cluster(ClusterEvent::MessageSkipped),
             b"cluster.message-invalid" => EventType::Cluster(ClusterEvent::MessageInvalid),
             b"cluster.node-id-renewed" => EventType::Cluster(ClusterEvent::NodeIdRenewed),
+            // inbuxa: coordinator connection
+            b"cluster.coordinator-connected" => EventType::Cluster(ClusterEvent::CoordinatorConnected),
+            b"cluster.coordinator-disconnected" => EventType::Cluster(ClusterEvent::CoordinatorDisconnected),
+            b"cluster.coordinator-error" => EventType::Cluster(ClusterEvent::CoordinatorError),
             b"dane.authentication-success" => EventType::Dane(DaneEvent::AuthenticationSuccess),
             b"dane.authentication-failure" => EventType::Dane(DaneEvent::AuthenticationFailure),
             b"dane.no-certificates-found" => EventType::Dane(DaneEvent::NoCertificatesFound),
@@ -742,6 +746,14 @@ impl EventType {
             EventType::Cluster(ClusterEvent::MessageSkipped) => "cluster.message-skipped",
             EventType::Cluster(ClusterEvent::MessageInvalid) => "cluster.message-invalid",
             EventType::Cluster(ClusterEvent::NodeIdRenewed) => "cluster.node-id-renewed",
+            // inbuxa: coordinator connection
+            EventType::Cluster(ClusterEvent::CoordinatorConnected) => {
+                "cluster.coordinator-connected"
+            }
+            EventType::Cluster(ClusterEvent::CoordinatorDisconnected) => {
+                "cluster.coordinator-disconnected"
+            }
+            EventType::Cluster(ClusterEvent::CoordinatorError) => "cluster.coordinator-error",
             EventType::Dane(DaneEvent::AuthenticationSuccess) => "dane.authentication-success",
             EventType::Dane(DaneEvent::AuthenticationFailure) => "dane.authentication-failure",
             EventType::Dane(DaneEvent::NoCertificatesFound) => "dane.no-certificates-found",
@@ -1524,6 +1536,10 @@ impl EventType {
             EventType::Cluster(ClusterEvent::MessageSkipped) => 47,
             EventType::Cluster(ClusterEvent::MessageInvalid) => 49,
             EventType::Cluster(ClusterEvent::NodeIdRenewed) => 275,
+            // inbuxa: coordinator connection
+            EventType::Cluster(ClusterEvent::CoordinatorConnected) => 644,
+            EventType::Cluster(ClusterEvent::CoordinatorDisconnected) => 645,
+            EventType::Cluster(ClusterEvent::CoordinatorError) => 646,
             EventType::Dane(DaneEvent::AuthenticationSuccess) => 67,
             EventType::Dane(DaneEvent::AuthenticationFailure) => 66,
             EventType::Dane(DaneEvent::NoCertificatesFound) => 69,
@@ -2176,6 +2192,10 @@ impl EventType {
             47 => Some(EventType::Cluster(ClusterEvent::MessageSkipped)),
             49 => Some(EventType::Cluster(ClusterEvent::MessageInvalid)),
             275 => Some(EventType::Cluster(ClusterEvent::NodeIdRenewed)),
+            // inbuxa: coordinator connection
+            644 => Some(EventType::Cluster(ClusterEvent::CoordinatorConnected)),
+            645 => Some(EventType::Cluster(ClusterEvent::CoordinatorDisconnected)),
+            646 => Some(EventType::Cluster(ClusterEvent::CoordinatorError)),
             67 => Some(EventType::Dane(DaneEvent::AuthenticationSuccess)),
             66 => Some(EventType::Dane(DaneEvent::AuthenticationFailure)),
             69 => Some(EventType::Dane(DaneEvent::NoCertificatesFound)),
@@ -3114,6 +3134,10 @@ impl EventType {
             EventType::Auth(AuthEvent::TooManyAttempts) => Level::Warn,
             EventType::Calendar(CalendarEvent::AlarmFailed) => Level::Warn,
             EventType::Cluster(ClusterEvent::SubscriberDisconnected) => Level::Warn,
+            // inbuxa: coordinator connection
+            EventType::Cluster(ClusterEvent::CoordinatorConnected) => Level::Info,
+            EventType::Cluster(ClusterEvent::CoordinatorDisconnected) => Level::Warn,
+            EventType::Cluster(ClusterEvent::CoordinatorError) => Level::Warn,
             EventType::Delivery(DeliveryEvent::MissingOutboundHostname) => Level::Warn,
             EventType::Delivery(DeliveryEvent::ConcurrencyLimitExceeded) => Level::Warn,
             EventType::Delivery(DeliveryEvent::RateLimitExceeded) => Level::Warn,
@@ -3244,6 +3268,10 @@ impl EventType {
             EventType::Cluster(ClusterEvent::MessageSkipped) => "PubSub message skipped",
             EventType::Cluster(ClusterEvent::MessageInvalid) => "Invalid PubSub message",
             EventType::Cluster(ClusterEvent::NodeIdRenewed) => "Node ID renewed",
+            // inbuxa: coordinator connection
+            EventType::Cluster(ClusterEvent::CoordinatorConnected) => "Coordinator connected",
+            EventType::Cluster(ClusterEvent::CoordinatorDisconnected) => "Coordinator unavailable",
+            EventType::Cluster(ClusterEvent::CoordinatorError) => "Coordinator error",
             EventType::Dane(DaneEvent::AuthenticationSuccess) => "DANE authentication successful",
             EventType::Dane(DaneEvent::AuthenticationFailure) => "DANE authentication failed",
             EventType::Dane(DaneEvent::NoCertificatesFound) => "No certificates found for DANE",
@@ -4322,6 +4350,10 @@ impl EventType {
             EventType::Cluster(ClusterEvent::MessageSkipped),
             EventType::Cluster(ClusterEvent::MessageInvalid),
             EventType::Cluster(ClusterEvent::NodeIdRenewed),
+            // inbuxa: coordinator connection
+            EventType::Cluster(ClusterEvent::CoordinatorConnected),
+            EventType::Cluster(ClusterEvent::CoordinatorDisconnected),
+            EventType::Cluster(ClusterEvent::CoordinatorError),
             EventType::Dane(DaneEvent::AuthenticationSuccess),
             EventType::Dane(DaneEvent::AuthenticationFailure),
             EventType::Dane(DaneEvent::NoCertificatesFound),

resources/schema/schema.json.sha256

@@ -1 +1 @@
-VbnFuwCOTBh0s2T-NuRhb2JaJr8Jl5s3LgXv4Pv2sTg
+XFI3xuKC_rH1KZyaVBF0uTIiRDXRqyYboijquiGz2eg

tests/src/cluster/coordinator.rs

@@ -0,0 +1,145 @@
+/*
+ * SPDX-FileCopyrightText: 2026 Coffey Labs
+ *
+ * SPDX-License-Identifier: AGPL-3.0-only
+ */
+
+//! A node that starts while its NATS coordinator is down joins the cluster
+//! once NATS comes up, without a restart, and reports the coordinator's
+//! connection on `/healthz/cluster` as it goes and comes back.
+
+use crate::utils::server::TestServerBuilder;
+use coordinator::Coordinator;
+use registry::{
+    schema::{
+        enums::NetworkListenerProtocol,
+        structs::{Coordinator as CoordinatorSetting, NatsCoordinator},
+    },
+    types::map::Map,
+};
+use serde_json::{Value, json};
+use std::time::{Duration, Instant};
+use testcontainers::{
+    GenericImage, ImageExt, core::IntoContainerPort, core::WaitFor, runners::AsyncRunner,
+};
+
+const HTTP_PORT: u16 = 11_310;
+const TOPIC: &str = "inbuxa-coordinator-test";
+
+#[tokio::test(flavor = "multi_thread")]
+pub async fn coordinator_reconnect_tests() {
+    println!("Running coordinator reconnect tests...");
+
+    // A port with no NATS server behind it, yet
+    let nats_port = std::net::TcpListener::bind("127.0.0.1:0")
+        .unwrap()
+        .local_addr()
+        .unwrap()
+        .port();
+    let config = NatsCoordinator {
+        addresses: Map::new(vec![format!("127.0.0.1:{nats_port}")]),
+        use_tls: false,
+        timeout_connection: 1_000u64.into(),
+        ..Default::default()
+    };
+
+    // 1. The node starts, without a build error, while NATS is down, and
+    // says so
+    let test = TestServerBuilder::new("coordinator_reconnect_tests")
+        .await
+        .with_object(CoordinatorSetting::Nats(config.clone()))
+        .await
+        .with_listener(NetworkListenerProtocol::Http, "http", HTTP_PORT, true)
+        .await
+        .build()
+        .await;
+    let coordinator = test.server.core.storage.coordinator.clone();
+    assert!(
+        coordinator.is_enabled(),
+        "a coordinator, though not connected"
+    );
+    assert_eq!(coordinator.is_connected(), Some(false));
+    assert_eq!(
+        cluster_health().await,
+        (503, json!({"coordinator": "disconnected"}))
+    );
+
+    // A subscription made now, as the broadcast subscriber makes it at
+    // startup, has to work once NATS is up
+    let mut stream = coordinator.subscribe(TOPIC).await.unwrap();
+
+    // 2. NATS comes up: the node connects on its own
+    let nats = GenericImage::new("nats", "latest")
+        .with_wait_for(WaitFor::message_on_stderr("Server is ready"))
+        .with_mapped_port(nats_port, 4222.tcp())
+        .start()
+        .await
+        .expect("Failed to start NATS container");
+    wait_for_health(200, "connected").await;
+    let other_node = coordinator::backend::nats::NatsPubSub::open(config.clone())
+        .await
+        .unwrap();
+    wait_until_connected(&other_node).await;
+    round_trip(&other_node, &mut stream, b"after startup").await;
+
+    // 3. NATS goes away: the node reports it; and it comes back: the node
+    // reconnects and the same subscription carries on
+    nats.stop().await.unwrap();
+    wait_for_health(503, "disconnected").await;
+    nats.start().await.unwrap();
+    wait_for_health(200, "connected").await;
+    wait_until_connected(&other_node).await;
+    round_trip(&other_node, &mut stream, b"after reconnect").await;
+
+    drop(nats);
+    if test.is_reset() {
+        test.temp_dir.delete();
+    }
+}
+
+async fn cluster_health() -> (u16, Value) {
+    let response = reqwest::Client::builder()
+        .danger_accept_invalid_certs(true)
+        .timeout(Duration::from_secs(5))
+        .build()
+        .unwrap()
+        .get(format!("https://127.0.0.1:{HTTP_PORT}/healthz/cluster"))
+        .send()
+        .await
+        .unwrap();
+    let status = response.status().as_u16();
+    (status, response.json().await.unwrap())
+}
+
+async fn wait_for_health(status: u16, state: &str) {
+    let started = Instant::now();
+    loop {
+        let health = cluster_health().await;
+        if health == (status, json!({"coordinator": state})) {
+            return;
+        }
+        assert!(
+            started.elapsed() < Duration::from_secs(30),
+            "expected {status} {state}, still {health:?}"
+        );
+        tokio::time::sleep(Duration::from_millis(250)).await;
+    }
+}
+
+async fn wait_until_connected(coordinator: &Coordinator) {
+    let started = Instant::now();
+    while coordinator.is_connected() != Some(true) {
+        assert!(started.elapsed() < Duration::from_secs(30), "not connected");
+        tokio::time::sleep(Duration::from_millis(100)).await;
+    }
+}
+
+/// Another node publishes; this one's subscription receives it.
+async fn round_trip(from: &Coordinator, stream: &mut coordinator::PubSubStream, payload: &[u8]) {
+    from.publish(TOPIC, payload.to_vec()).await.unwrap();
+    let message = tokio::time::timeout(Duration::from_secs(10), stream.next())
+        .await
+        .expect("no message within 10 seconds")
+        .expect("subscription ended");
+    assert_eq!(message.payload(), payload);
+}

tests/src/cluster/mod.rs

@@ -2,7 +2,11 @@
  * SPDX-FileCopyrightText: 2020 Stalwart Labs LLC <[email protected]>
  *
  * SPDX-License-Identifier: AGPL-3.0-only OR LicenseRef-SEL
+ *
+ * Modified by Coffey Labs in 2026 for INBUXA.
  */
 
 pub mod broadcast;
+#[cfg(feature = "nats")]
+pub mod coordinator; // inbuxa: coordinator reconnects
 pub mod stress;

tests/src/smtp/inbound/milter.rs

@@ -2,6 +2,8 @@
  * SPDX-FileCopyrightText: 2020 Stalwart Labs LLC <[email protected]>
  *
  * SPDX-License-Identifier: AGPL-3.0-only OR LicenseRef-SEL
+ *
+ * Modified by Coffey Labs in 2026 for INBUXA.
  */
 
 use crate::{
@@ -75,7 +77,7 @@ async fn milter_session() {
                 else_: "true".into(),
                 ..Default::default()
             },
-            hostname: "127.0.0.1".into(),
+            hostname: "localhost".into(), // inbuxa: resolved when the session connects
             port: 9332,
             use_tls: false,
             stages: Map::new(vec![MtaStage::Data]),

tests/src/store/mod.rs

@@ -21,6 +21,7 @@ pub mod replica_cluster; // inbuxa: read replicas across nodes
 pub mod scaleout; // inbuxa: scale-out storage
 #[cfg(any(feature = "postgres", feature = "mysql"))]
 pub mod sql_timeout;
+pub mod task_locks; // inbuxa: task locks across nodes
 
 use crate::utils::server::TestServerBuilder;
 use std::io::Read;

tests/src/store/query.rs

@@ -128,6 +128,11 @@ pub async fn test(test: &TestServer) {
     println!("Running trace document tests...");
     test_trace_documents(store.clone()).await;
 
+    // inbuxa: address fields match by full address, local part, domain and
+    // display name on every backend
+    println!("Running address search tests...");
+    test_address_search(store.clone()).await;
+
     // Large document insert test
     println!("Running large document insert tests...");
     let mut large_text = String::with_capacity(20 * 1024 * 1024);
@@ -972,3 +977,155 @@ async fn test_trace_documents(store: SearchStore) {
             .unwrap();
     }
 }
+
+// inbuxa: the message indexer passes each display name and each address of
+// From/To/Cc/Bcc as keyword text (Language::None). The built-in index splits
+// that text into words, so an address is found by its full form, its local
+// part, its domain or a display-name word; PostgreSQL kept the whole address
+// as one token and MySQL dropped stopwords such as "com" and words under three
+// characters. The expected results below are the built-in (RocksDB/SQLite)
+// results and must be the same on every backend.
+async fn test_address_search(store: SearchStore) {
+    const ACCOUNT_ID: u32 = 7;
+    let messages: [[&[(&str, &str)]; 4]; 5] = [
+        // From, To, Cc, Bcc
+        [
+            &[("Amazon.com", "[email protected]")],
+            &[("Jane Doe", "[email protected]")],
+            &[],
+            &[],
+        ],
+        [
+            &[("", "[email protected]")],
+            &[("", "[email protected]")],
+            &[("Jane Doe", "[email protected]")],
+            &[],
+        ],
+        [
+            &[("GitHub", "[email protected]")],
+            &[("Jo Li", "[email protected]")],
+            &[],
+            &[("Audit", "[email protected]")],
+        ],
+        [
+            &[("Jane Doe", "[email protected]")],
+            &[("Amazon Web Services", "[email protected]")],
+            &[("Bob", "[email protected]")],
+            &[("", "[email protected]")],
+        ],
+        [
+            &[("Newsletter", "[email protected]")],
+            &[("", "[email protected]")],
+            &[],
+            &[],
+        ],
+    ];
+    let fields = [
+        EmailSearchField::From,
+        EmailSearchField::To,
+        EmailSearchField::Cc,
+        EmailSearchField::Bcc,
+    ];
+
+    let mut documents = Vec::new();
+    let mut mask = RoaringBitmap::new();
+    for (document_id, message) in messages.iter().enumerate() {
+        let mut document = IndexDocument::new(SearchIndex::Email)
+            .with_account_id(ACCOUNT_ID)
+            .with_document_id(document_id as u32);
+        for (field, addresses) in fields.iter().zip(message.iter()) {
+            for (name, address) in addresses.iter() {
+                if !name.is_empty() {
+                    document.index_text(field.clone(), name, Language::None);
+                }
+                document.index_text(field.clone(), address, Language::None);
+            }
+        }
+        document.index_unsigned(EmailSearchField::ReceivedAt, document_id as u64);
+        documents.push(document);
+        mask.insert(document_id as u32);
+    }
+    store.index(documents).await.unwrap();
+    if let SearchStore::ElasticSearch(store) = &store {
+        store.refresh_index(SearchIndex::Email).await.unwrap();
+    }
+
+    for (field, text, expected) in [
+        // full address
+        (EmailSearchField::From, "[email protected]", vec![0u32]),
+        (EmailSearchField::To, "[email protected]", vec![0]),
+        (EmailSearchField::Cc, "[email protected]", vec![1]),
+        (EmailSearchField::Bcc, "[email protected]", vec![3]),
+        (EmailSearchField::To, "[email protected]", vec![1, 2]),
+        // local part
+        (EmailSearchField::From, "noreply", vec![0, 2]),
+        (EmailSearchField::To, "jo", vec![1, 2]),
+        (EmailSearchField::Cc, "bob", vec![3]),
+        (EmailSearchField::Bcc, "audit", vec![2]),
+        // domain
+        (EmailSearchField::From, "amazon.com", vec![0, 1]),
+        (EmailSearchField::From, "amazon", vec![0, 1]),
+        (EmailSearchField::To, "example.org", vec![0, 4]),
+        (EmailSearchField::To, "io.de", vec![1, 2]),
+        (EmailSearchField::Cc, "example.net", vec![3]),
+        (EmailSearchField::Bcc, "example.org", vec![2]),
+        (EmailSearchField::From, "www.example.com", vec![4]),
+        (EmailSearchField::From, "com", vec![0, 1, 2, 4]),
+        // display name
+        (EmailSearchField::From, "Jane", vec![3]),
+        (EmailSearchField::From, "jane doe", vec![3]),
+        (EmailSearchField::To, "Web Services", vec![3]),
+        (EmailSearchField::To, "Li", vec![2]),
+        (EmailSearchField::Cc, "Doe", vec![1]),
+        (EmailSearchField::Bcc, "Audit", vec![2]),
+        // hyphenated local part
+        (EmailSearchField::From, "shipment-tracking", vec![1]),
+        (EmailSearchField::From, "tracking", vec![1]),
+        // no match
+        (EmailSearchField::From, "amazon.org", vec![]),
+        (EmailSearchField::To, "noreply", vec![]),
+        (EmailSearchField::Bcc, "jane", vec![]),
+    ] {
+        let ids = store
+            .query_account(
+                SearchQuery::new(SearchIndex::Email)
+                    .with_filters(vec![
+                        SearchFilter::eq(SearchField::AccountId, ACCOUNT_ID),
+                        SearchFilter::has_keyword(field.clone(), text),
+                    ])
+                    .with_comparator(SearchComparator::ascending(EmailSearchField::ReceivedAt))
+                    .with_mask(mask.clone()),
+            )
+            .await
+            .unwrap();
+        assert_eq!(ids, expected, "{field:?} {text:?}");
+    }
+
+    // TEXT-style search across all address fields
+    let ids = store
+        .query_account(
+            SearchQuery::new(SearchIndex::Email)
+                .with_filters(vec![
+                    SearchFilter::eq(SearchField::AccountId, ACCOUNT_ID),
+                    SearchFilter::Or,
+                    SearchFilter::has_keyword(EmailSearchField::From, "example.org"),
+                    SearchFilter::has_keyword(EmailSearchField::To, "example.org"),
+                    SearchFilter::has_keyword(EmailSearchField::Cc, "example.org"),
+                    SearchFilter::has_keyword(EmailSearchField::Bcc, "example.org"),
+                    SearchFilter::End,
+                ])
+                .with_comparator(SearchComparator::ascending(EmailSearchField::ReceivedAt))
+                .with_mask(mask.clone()),
+        )
+        .await
+        .unwrap();
+    assert_eq!(ids, vec![0, 1, 2, 3, 4]);
+
+    store
+        .unindex(
+            SearchQuery::new(SearchIndex::Email)
+                .with_filter(SearchFilter::eq(SearchField::AccountId, ACCOUNT_ID)),
+        )
+        .await
+        .unwrap();
+}

tests/src/store/task_locks.rs

@@ -0,0 +1,184 @@
+/*
+ * SPDX-FileCopyrightText: 2026 Coffey Labs
+ *
+ * SPDX-License-Identifier: AGPL-3.0-only
+ */
+
+//! Task locks across nodes: tasks claimed by a node that then disappears
+//! run elsewhere once its locks expire, and a node that stops gracefully
+//! hands its locks back at once. The other node is played by writing its
+//! locks straight into the shared in-memory store, as a node that claimed
+//! the tasks and died leaves them.
+
+use crate::utils::server::TestServerBuilder;
+use common::{KV_LOCK_TASK, Server};
+use registry::schema::{
+    enums::IndexDocumentType,
+    structs::{Task, TaskIndexDocument, TaskStatus},
+};
+use services::task_manager::lock::{TaskLockManager, release_task_locks};
+use std::time::{Duration, Instant};
+use store::{
+    ValueKey,
+    write::{BatchBuilder, TaskQueueClass, ValueClass},
+};
+use utils::snowflake::SnowflakeIdGenerator;
+
+// Short enough for a test, long enough that the recheck interval (a twelfth
+// of it) is well below it
+const LOCK_EXPIRY: u64 = 12;
+
+#[tokio::test(flavor = "multi_thread")]
+pub async fn task_lock_tests() {
+    let test = TestServerBuilder::new("task_lock_tests")
+        .await
+        .build()
+        .await;
+    let server = test.server.clone();
+    println!(
+        "Running task lock tests on {}...",
+        std::env::var("STORE").unwrap_or_default()
+    );
+    server.inner.ipc.task_locks.set_expiry(LOCK_EXPIRY);
+
+    // 1. Another node claimed the tasks and died. Its locks block them until
+    // they expire; then this node runs them, without waiting for anything
+    // else to wake it
+    let ids = new_task_ids(4);
+    for id in &ids {
+        assert!(foreign_lock(&server, *id, LOCK_EXPIRY).await);
+    }
+    schedule(&server, &ids).await;
+    let started = Instant::now();
+    server.notify_task_queue();
+    tokio::time::sleep(Duration::from_secs(3)).await;
+    assert_eq!(pending(&server, &ids).await, ids.len(), "held by the other node");
+    wait_until_done(&server, &ids, Duration::from_secs(LOCK_EXPIRY + 10)).await;
+    let elapsed = started.elapsed();
+    assert!(
+        elapsed >= Duration::from_secs(LOCK_EXPIRY - 2),
+        "ran before the other node's locks expired: {elapsed:?}"
+    );
+
+    // 2. The other node's locks outlive what this node expects: claimed just
+    // after this node looked, or by a node whose clock runs ahead. This node
+    // keeps checking at the recheck interval, so the tasks run soon after
+    // those locks expire, not a whole lock lifetime later
+    let held_for = LOCK_EXPIRY + LOCK_EXPIRY / 2;
+    let ids = new_task_ids(4);
+    for id in &ids {
+        assert!(foreign_lock(&server, *id, held_for).await);
+    }
+    schedule(&server, &ids).await;
+    let started = Instant::now();
+    server.notify_task_queue();
+    wait_until_done(&server, &ids, Duration::from_secs(held_for + 8)).await;
+    let elapsed = started.elapsed();
+    assert!(
+        elapsed >= Duration::from_secs(held_for - 2),
+        "ran before the other node's locks expired: {elapsed:?}"
+    );
+
+    // 3. A graceful stop releases the locks this node holds: another node
+    // can claim those tasks at once, and this one claims nothing more
+    let ids = new_task_ids(3);
+    for id in &ids {
+        assert!(server.try_lock_task(*id).await, "claim {id}");
+    }
+    assert_eq!(server.inner.ipc.task_locks.held(), ids.len());
+    for id in &ids {
+        assert!(
+            !foreign_lock(&server, *id, LOCK_EXPIRY).await,
+            "held while this node runs"
+        );
+    }
+    assert_eq!(release_task_locks(&server).await, ids.len());
+    assert_eq!(server.inner.ipc.task_locks.held(), 0);
+    for id in &ids {
+        assert!(
+            foreign_lock(&server, *id, LOCK_EXPIRY).await,
+            "released on stop: {id}"
+        );
+    }
+    let [id] = new_task_ids(1)[..] else {
+        unreachable!()
+    };
+    assert!(!server.try_lock_task(id).await, "a stopping node claims nothing");
+
+    for id in ids {
+        let _ = server
+            .in_memory_store()
+            .remove_lock(KV_LOCK_TASK, &id.to_be_bytes())
+            .await;
+    }
+    if test.is_reset() {
+        test.temp_dir.delete();
+    }
+}
+
+fn new_task_ids(count: usize) -> Vec<u64> {
+    (0..count)
+        .map(|_| SnowflakeIdGenerator::global_id().unwrap())
+        .collect()
+}
+
+/// The other node's claim on a task, as its task manager takes it.
+async fn foreign_lock(server: &Server, id: u64, seconds: u64) -> bool {
+    server
+        .in_memory_store()
+        .try_lock(KV_LOCK_TASK, &id.to_be_bytes(), seconds)
+        .await
+        .unwrap()
+}
+
+/// Unindex tasks for files that don't exist: files aren't search-indexed and
+/// there is no undelete note, so running one only drops it from the queue.
+async fn schedule(server: &Server, ids: &[u64]) {
+    let mut batch = BatchBuilder::new();
+    for (n, id) in ids.iter().enumerate() {
+        batch.schedule_task_with_id(
+            *id,
+            Task::UnindexDocument(TaskIndexDocument {
+                account_id: 0u32.into(),
+                document_id: (u32::MAX - n as u32).into(),
+                document_type: IndexDocumentType::File,
+                status: TaskStatus::now(),
+            }),
+        );
+    }
+    server.store().write(batch.build_all()).await.unwrap();
+}
+
+async fn pending(server: &Server, ids: &[u64]) -> usize {
+    let mut count = 0;
+    for id in ids {
+        if server
+            .store()
+            .get_value::<Task>(ValueKey::from(ValueClass::TaskQueue(
+                TaskQueueClass::Task { id: *id },
+            )))
+            .await
+            .unwrap()
+            .is_some()
+        {
+            count += 1;
+        }
+    }
+    count
+}
+
+async fn wait_until_done(server: &Server, ids: &[u64], within: Duration) {
+    let started = Instant::now();
+    loop {
+        let left = pending(server, ids).await;
+        if left == 0 {
+            return;
+        }
+        assert!(
+            started.elapsed() < within,
+            "{left} task(s) still pending after {:?}",
+            started.elapsed()
+        );
+        tokio::time::sleep(Duration::from_millis(250)).await;
+    }
+}

tests/src/system/mod.rs

@@ -20,6 +20,7 @@ pub mod monitoring;
 pub mod oidc;
 pub mod purge;
 pub mod quota;
+pub mod reload; // inbuxa: reloads and build errors
 pub mod security;
 pub mod task;
 pub mod tenant;

tests/src/system/reload.rs

@@ -0,0 +1,213 @@
+/*
+ * SPDX-FileCopyrightText: 2026 Coffey Labs
+ *
+ * SPDX-License-Identifier: AGPL-3.0-only
+ */
+
+// inbuxa: a settings reload isn't held back by a DNS lookup, or by objects
+// that already failed when the running settings were built; an error in an
+// object that built then still refuses it, and says which object.
+
+use crate::utils::server::{TestServer, TestServerBuilder};
+use common::{BuildServer, config::mailstore::spamfilter::PyzorConfig, ipc::RegistryChange};
+use registry::schema::{
+    enums::TracingLevel,
+    prelude::{ObjectType, Property},
+    structs::{Action, Expression, MtaStageAuth, SpamPyzor, Tracer, TracerStdout},
+};
+
+#[tokio::test(flavor = "multi_thread")]
+pub async fn reload_tests() {
+    let mut test = TestServerBuilder::new("reload_tests")
+        .await
+        .with_default_listeners()
+        .await
+        .with_object(MtaStageAuth {
+            require: Expression {
+                else_: "false".to_string(),
+                ..Default::default()
+            },
+            ..Default::default()
+        })
+        .await
+        .build()
+        .await;
+
+    let admin = test
+        .create_user_account(
+            "admin",
+            "[email protected]",
+            "these_pretzels_are_making_me_thirsty",
+            &[],
+            "Admin",
+        )
+        .await;
+    test.account("admin")
+        .assign_roles_to_account(admin.id(), &["user", "system"])
+        .await;
+    test.insert_account(admin);
+
+    test_unresolvable_pyzor(&test).await;
+    test_build_errors(&test).await;
+
+    if test.is_reset() {
+        test.temp_dir.delete();
+    }
+}
+
+async fn test_unresolvable_pyzor(test: &TestServer) {
+    println!("Running reload with an unresolvable Pyzor host...");
+    let admin = test.account("[email protected]");
+
+    // Upstream resolved the host while building the settings and refused the
+    // reload when that failed.
+    admin
+        .registry_update_setting(
+            SpamPyzor {
+                enable: true,
+                host: "pyzor.invalid".into(),
+                port: 24441,
+                ..Default::default()
+            },
+            &[Property::Enable, Property::Host, Property::Port],
+        )
+        .await;
+    admin.reload_settings().await;
+
+    let pyzor = running_pyzor(test);
+    assert_eq!(pyzor.host, "pyzor.invalid");
+    assert_eq!(pyzor.port, 24441);
+    assert!(pyzor.address().await.is_err());
+
+    // An IP address needs no lookup
+    admin
+        .registry_update_setting(
+            SpamPyzor {
+                host: "192.0.2.1".into(),
+                ..Default::default()
+            },
+            &[Property::Host],
+        )
+        .await;
+    admin.reload_settings().await;
+    assert_eq!(
+        running_pyzor(test).address().await.unwrap().to_string(),
+        "192.0.2.1:24441"
+    );
+}
+
+async fn test_build_errors(test: &TestServer) {
+    println!("Running reload with build errors...");
+    let admin = test.account("[email protected]");
+    let pyzor_ratio = running_pyzor(test).ratio;
+    assert_ne!(pyzor_ratio, 0.25);
+
+    // Two console tracers: only one is allowed, so the build of one of them
+    // fails. Neither existed when the running settings were built.
+    let mut tracer_ids = Vec::new();
+    for _ in 0..2 {
+        tracer_ids.push(
+            admin
+                .registry_create_object(Tracer::Stdout(TracerStdout {
+                    enable: true,
+                    level: TracingLevel::Error,
+                    ..Default::default()
+                }))
+                .await,
+        );
+    }
+    admin
+        .registry_update_setting(
+            SpamPyzor {
+                ratio: 0.25.into(),
+                ..Default::default()
+            },
+            &[Property::Ratio],
+        )
+        .await;
+
+    // A new error refuses the reload and names the object
+    let err = admin
+        .registry_create_object_expect_err(Action::ReloadSettings)
+        .await;
+    let description = err.description.clone().unwrap_or_default();
+    assert!(
+        description.starts_with("Settings were not reloaded. ")
+            && description.contains("Tracer")
+            && description.contains("Only one console tracer is allowed"),
+        "{err:?}"
+    );
+    assert_eq!(running_pyzor(test).ratio, pyzor_ratio);
+
+    // Had the running settings been built with that tracer failing, as a
+    // restart now would, the same error doesn't hold the reload back.
+    let result = Box::pin(
+        test.server
+            .reload_registry(RegistryChange::Reload(ObjectType::DataStore)),
+    )
+    .await
+    .unwrap();
+    assert!(!result.replaced_core);
+    assert_eq!(result.errors.len(), 1, "{:?}", result.errors);
+    test.server.record_build_errors(&result.errors);
+
+    admin.reload_settings().await;
+    assert_eq!(running_pyzor(test).ratio, 0.25);
+    let result = Box::pin(
+        test.server
+            .reload_registry(RegistryChange::Reload(ObjectType::DataStore)),
+    )
+    .await
+    .unwrap();
+    assert!(result.replaced_core);
+    assert!(result.errors.is_empty());
+    assert_eq!(result.known_errors.len(), 1);
+
+    // Once fixed, the object is no longer known to fail, so a new error
+    // there refuses the reload again.
+    admin
+        .registry_destroy(ObjectType::Tracer, tracer_ids.iter())
+        .await
+        .assert_destroyed(&tracer_ids);
+    admin.reload_settings().await;
+    let result = Box::pin(
+        test.server
+            .reload_registry(RegistryChange::Reload(ObjectType::DataStore)),
+    )
+    .await
+    .unwrap();
+    assert!(result.replaced_core);
+    assert!(result.errors.is_empty() && result.known_errors.is_empty());
+
+    for _ in 0..2 {
+        tracer_ids.push(
+            admin
+                .registry_create_object(Tracer::Stdout(TracerStdout {
+                    enable: true,
+                    level: TracingLevel::Error,
+                    ..Default::default()
+                }))
+                .await,
+        );
+    }
+    admin
+        .registry_create_object_expect_err(Action::ReloadSettings)
+        .await;
+    let tracer_ids = tracer_ids.split_off(2);
+    admin
+        .registry_destroy(ObjectType::Tracer, tracer_ids.iter())
+        .await
+        .assert_destroyed(&tracer_ids);
+    admin.reload_settings().await;
+}
+
+fn running_pyzor(test: &TestServer) -> PyzorConfig {
+    test.server
+        .inner
+        .build_server()
+        .core
+        .spam
+        .pyzor
+        .clone()
+        .expect("Pyzor enabled")
+}