This avoids having the preemptive refresh probability start off too high. For example, LoadMonitor and LoadBalancer::isMasterRunningReadOnly() are effected by this.
Description
Details
| Subject | Repo | Branch | Lines +/- | |
|---|---|---|---|---|
| objectcache: fix WANCache getWithSetCallback() when TTL < "lowTTL" | mediawiki/core | master | +130 -81 |
Related Objects
Event Timeline
Change 634116 had a related patch set uploaded (by Aaron Schulz; owner: Aaron Schulz):
[mediawiki/core@master] objectcache: fix WANCache getWithSetCallback() when TTL < "lowTTL"
This avoids having the preemptive refresh probability start off too high. For example, LoadMonitor and LoadBalancer::isMasterRunningReadOnly() are effected by this.
I assume this is related to T252564#6212081, right? If so, could you describe in a bit more detail how this might lead to unexpected read-only mode? The busyValue is set to something that afaik assumes neutral loads in read-write (with hypothetical fallback to late-discovery of a hardware read-only mode). So where is the read-only confusion coming from?
It could be related if the problem has to do with regenerations (since it was intermitted). In any case, excessive regeneration is a problem in itself.
Change 634116 merged by jenkins-bot:
[mediawiki/core@master] objectcache: fix WANCache getWithSetCallback() when TTL < "lowTTL"