Renaming a server leaves its old node behind as "not responding" for a day #458

Closed
opened 2026-11-18 20:38:54 +00:00 by jcoffey-dev · 1 comment
Owner

Reported in the forum (t/324): after renaming a single server to a new mail host, the dashboard showed "2 nodes not responding" while mail and the console worked fine.

The installer sets the container's hostname to the mail host, and the server keys its node lease by HOSTNAME (crates/store/src/registry/local.rs). After a rename the server takes a new lease under the new name. The old one stops renewing, shows as Stale after three minutes, and stays until it goes inactive 24 hours later (DEAD_NODE_TIMEOUT). ClusterNode can't be deleted from the console, so there's nothing the admin can do but wait.

Fix: each node remembers the hostname it last leased under, in a small file next to its local config. When that name changes, the node releases the old name's lease once it has stopped renewing. A node that really is still running under the old name keeps renewing, so its lease is left alone.

Reported in the forum (t/324): after renaming a single server to a new mail host, the dashboard showed "2 nodes not responding" while mail and the console worked fine. The installer sets the container's hostname to the mail host, and the server keys its node lease by `HOSTNAME` (`crates/store/src/registry/local.rs`). After a rename the server takes a new lease under the new name. The old one stops renewing, shows as Stale after three minutes, and stays until it goes inactive 24 hours later (`DEAD_NODE_TIMEOUT`). ClusterNode can't be deleted from the console, so there's nothing the admin can do but wait. Fix: each node remembers the hostname it last leased under, in a small file next to its local config. When that name changes, the node releases the old name's lease once it has stopped renewing. A node that really is still running under the old name keeps renewing, so its lease is left alone.
jcoffey-dev added this to the 27.1 milestone 2026-11-18 20:38:54 +00:00
jcoffey-dev added the bug label 2026-11-18 20:38:54 +00:00
jcoffey-dev self-assigned this 2026-11-18 20:38:54 +00:00
Author
Owner

Fixed by #460. Each server now remembers the hostname it last ran under (node-hostname in /etc/inbuxa). After a rename it drops the old name's node record once that record has missed two check-ins, before the dashboard would show it as not responding. A server that's still checking in under the old name keeps its record.

Records left behind by renames before this fix still expire after 24 hours, as before. It'll be in the next release.

Fixed by #460. Each server now remembers the hostname it last ran under (`node-hostname` in `/etc/inbuxa`). After a rename it drops the old name's node record once that record has missed two check-ins, before the dashboard would show it as not responding. A server that's still checking in under the old name keeps its record. Records left behind by renames before this fix still expire after 24 hours, as before. It'll be in the next release.
Sign in to join this conversation.
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: inbuxa/inbuxa-server#458