Skip to content

Cluster Topology view shows stale nodes: 30s refresh broadcasts but never re-discovers #511

Description

@ravjotbrar

Summary

The Cluster Topology view reflects the topology captured at connect time and never updates. Nodes added or removed from the cluster afterward are not shown, even though a refresh loop runs every 30s.

Root cause

Regressed in #279. refreshAllClusterRegistries (the 30s loop) broadcasts clusterNodesRegistry to clients but no longer re-discovers topology first — the per-cycle updateClusterNodeRegistry call was removed. So it re-broadcasts the same stale snapshot forever.

The frontend Cluster Topology route depends on this broadcast for node topology (mount only fetches INFO/utilization stats, not the node list), so there is no other path that refreshes it.

Expected

The refresh loop should re-discover each tracked cluster's topology before broadcasting, so clients receive the current cluster nodes.

Notes

Each cluster must be re-discovered through a live client that belongs to it (topology discovery keys off the client's own CLUSTER SLOTS); clusters with no live client are left unchanged.

Activity

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions