Dashboard: Cluster Health replaces Server Memory in a cluster #17

Merged

jcoffey-dev merged 1 commits from feature/cluster-health into main

2026-09-25 08:08:21 +00:00

jcoffey-dev commented

2026-09-25 08:06:03 +00:00

Owner

When more than one node holds a lease, the overview's Server Memory card gives way to Cluster Health:

healthy nodes out of the total (3 / 3), with the healthy and unhealthy counts under it;
each silent node named, with how long it has been quiet;
a link to the node list when the viewer may open it.

It reads x:ClusterNode every 30 seconds. Healthy means Active. With the server's node heartbeat (inbuxa-server, feature/node-heartbeat) a node turns Stale after 3 minutes without a renewal; older servers take an hour. Inactive leases (a day silent) are left out. On a single server, or when the viewer can't list nodes, the dashboard is unchanged.

New strings use in-code English defaults, like the rest of the dashboard: cluster.health, cluster.healthy, cluster.unhealthy, cluster.quietMinutes, cluster.quietHours, cluster.quietUnknown, cluster.healthLink. en.json has no dashboard keys, so no catalogue changes.

Checks: typecheck, eslint and vitest (376 tests, 5 new) pass.

When more than one node holds a lease, the overview's **Server Memory** card gives way to **Cluster Health**: - healthy nodes out of the total (`3 / 3`), with the healthy and unhealthy counts under it; - each silent node named, with how long it has been quiet; - a link to the node list when the viewer may open it. It reads `x:ClusterNode` every 30 seconds. Healthy means Active. With the server's node heartbeat (inbuxa-server, feature/node-heartbeat) a node turns Stale after 3 minutes without a renewal; older servers take an hour. Inactive leases (a day silent) are left out. On a single server, or when the viewer can't list nodes, the dashboard is unchanged. New strings use in-code English defaults, like the rest of the dashboard: `cluster.health`, `cluster.healthy`, `cluster.unhealthy`, `cluster.quietMinutes`, `cluster.quietHours`, `cluster.quietUnknown`, `cluster.healthLink`. `en.json` has no dashboard keys, so no catalogue changes. Checks: typecheck, eslint and vitest (376 tests, 5 new) pass.

jcoffey-dev added 1 commit 2026-09-25 08:06:03 +00:00

Dashboard: Cluster Health replaces Server Memory in a cluster

ci / build (pull_request) Successful in 1m0s

Details

ci / publish (pull_request) Skipped

Details

8a41fef327

When more than one node holds a lease, the overview's Server Memory
card gives way to Cluster Health: healthy nodes out of the total, the
number unhealthy, and each silent node named with how long it has been
quiet. It reads x:ClusterNode every 30 seconds and links to the node
list when the viewer may open it.

Healthy means Active. A server with the node heartbeat marks a node
Stale after three minutes without a lease renewal; older servers take
an hour. Inactive leases (a day silent) are left out. On a single
server, or without permission to list nodes, the dashboard is
unchanged.

jcoffey-dev merged commit 645f4b22a0 into main

2026-09-25 08:08:21 +00:00

jcoffey-dev deleted branch feature/cluster-health

2026-09-25 08:08:21 +00:00

jcoffey-dev referenced this pull request

2026-09-26 21:21:30 +00:00

Version 2026.9.26 #19

jcoffey-dev referenced this pull request

2026-09-30 07:26:47 +00:00

ci: a cancelled GitHub run no longer reports failure to Gitea #67