You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
Commit 9c33f17
Browse filesBrowse the repository at this point in the historyBrowse files
fix(postgres): restart the plugin rollout whenever a backup job attaches it
Only the first attach after the latch restarted the rollout. In the
platform flow the chart renders no spec.plugins, so an in-place restore
recreates the Cluster without the plugin, and a backup job held through
that restore attaches it again once released. The rollout condition
then still carried its earlier start, the restored instances were newer
than it and ready, and the Backup was created in the same pass, before
CloudNativePG had rolled anything; such a Backup fails with "requested
plugin is not available" or "instance manager was restarted".
Every write of the job that finds the plugin missing now starts the
rollout over. A plugin attached by another writer is still timed from
the latch.
Assisted-by: LLM
Signed-off-by: Alexey Artamonov <[email protected]>
0 commit comments