Fix errors logged on standbys when we try to write versions to storage #13042
Conversation
| @@ -2000,6 +2000,9 @@ func (s standardUnsealStrategy) unseal(ctx context.Context, logger log.Logger, c | |||
| return err | |||
| } | |||
| } | |||
| if err := c.handleVersionTimeStamps(ctx); err != nil { | |||
There was a problem hiding this comment.
If this executes on a perf standby when the standby unseals, will it cause the standby to error on the storage write?
There was a problem hiding this comment.
I moved it here because standardUnsealStrategy doesn't apply to perf standbys.
There was a problem hiding this comment.
Do we want to load but not store the values on perf standbys?
There was a problem hiding this comment.
D'oh, good point. Will fix.
There was a problem hiding this comment.
Well, I guess I'll fix in ent once this is merged.
There was a problem hiding this comment.
Decided against loading on perf standbys since it wouldn't be reliable: #13081
| @@ -2157,11 +2160,6 @@ func (c *Core) postUnseal(ctx context.Context, ctxCancelFunc context.CancelFunc, | |||
| c.logger.Warn("post-unseal post seal migration failed", "error", err) | |||
| } | |||
| } | |||
| err := c.HandleVersionTimeStamps(c.activeContext) | |||
| if err != nil { | |||
| c.logger.Warn("post-unseal version timestamp setup failed", "error", err) | |||
There was a problem hiding this comment.
Is this the extra log line you were talking about in the PR description, where essentially the standby would always warn about the timestamp loading/storing not succeeding in the logs?
There was a problem hiding this comment.
I didn't actually test and see the log line myself, I speculated that we'd see this line on perf standbys.
I haven't actually seen error logs, but we know that only the active node should be allowed to write to storage.
While I was at it, I made some methods private that didn't need to be public and removed some dead code.