internal/natscluster, docs: observe through the monitoring port while servers do not all answer the system user #42

Merged
mikluko merged 1 commit from fix/observe-auth-switch into main 2026-10-05 10:00:55 +00:00
mikluko commented 2026-10-05 08:40:34 +00:00 (Migrated from github.com)

Servers still running a config without auth answer nothing on $SYS, so adding auth to a running NatsCluster held the rollout gate shut before the first restart. A $SYS failure is now masked whenever the monitoring port observes the cluster; bad systemCredentials surface only where a reload or scale-down needs $SYS.

Chosen over gating on "no system account running", which would restart every server without Settled.

Not run end to end on a real cluster; the tests cover observation, not the restart sequence.

Fixes #39

Servers still running a config without auth answer nothing on $SYS, so adding auth to a running NatsCluster held the rollout gate shut before the first restart. A $SYS failure is now masked whenever the monitoring port observes the cluster; bad systemCredentials surface only where a reload or scale-down needs $SYS. Chosen over gating on "no system account running", which would restart every server without Settled. Not run end to end on a real cluster; the tests cover observation, not the restart sequence. Fixes #39
Sign in to join this conversation.
No description provided.