litestream-oom-284
Human
#284 slice A: carlos-litestream's memory cap is sized from the database count, its restart loop backs off, and a per-unit restarts alert fires when a platform unit cannot stay up. Astra's fifth pass says MERGE-READY. Not merged: waiting on Paul's copy review (10 strings), a census ruling asked on #280, and his go-ahead.
Commits
13
Files changed
20
+1434 −26
Discussions
1
1 open
Tasks
3/9
done
Discussions 1 open
Tasks 3/9 done
- Slice A: size carlos-litestream MemoryMax from the database count Done
- Slice A: back off the litestream restart loop and stop it flooding the journal Done
- Slice A: box metric and alert for a restart-looping unit Done
- Moved to slice B (branch replica-freshness-284): alert for replica freshness Not started
- Slice B (branch replica-freshness-284): design; revs 1 and 2 rejected by Astra, mechanism change proposed on #290, needs Paul's ruling Started
Activity
- claude agent, operated by paul@keymail.dev replied to #284 slice B design: replica freshness for both replicators, and where it shows
- task updated: Slice A: Astra round 5 MERGE-READY, CI green on fa63c1ea; waiting on Paul's copy review, the census ruling and his go-ahead
- task updated: Slice A: copy review of the verb output, alert sentences and console alert text (Paul)
- task updated: Slice A: plan14 root-reads census overlay (written, eight rows); ruling asked of the census owner on #280
- claude agent, operated by paul@keymail.dev pushed fa63c1ead ops tests: the under-the-lock count test fails if the count moves before the lock (#284)
- claude agent, operated by paul@keymail.dev pushed 2ec2e90d4 ops litestream: the cap is sized from the config on disk, under the lock (#284)
- claude agent, operated by paul@keymail.dev pushed 01e126915 ops litestream: daemon-reload under a one-minute deadline (#284)
- claude agent, operated by paul@keymail.dev pushed 0fa3b84c1 memory drop-in under a lock; the restart watch keeps wall time and a present counter (#284)
- claude agent, operated by paul@keymail.dev pushed db57e5cae restarts watch and memory drop-in: the last three review findings (#284)
- task updated: Slice B (branch replica-freshness-284): design; revs 1 and 2 rejected by Astra, mechanism change proposed on #290, needs Paul's ruling
- claude agent, operated by paul@keymail.dev replied to #284 slice B design: replica freshness for both replicators, and where it shows
- claude agent, operated by paul@keymail.dev pushed bf7074d38 restarts alert: one rule per unit, a watch that survives resets, gaps and the agent (#284)
- claude agent, operated by paul@keymail.dev replied to #284 slice B design: replica freshness for both replicators, and where it shows
- claude agent, operated by paul@keymail.dev pushed 7cbde9452 census: overlay ls284 for the litestream cap, backoff and restart probe (#284)
- claude agent, operated by paul@keymail.dev pushed 1615e8605 Merge remote-tracking branch 'origin/main' into litestream-oom-284
- claude agent, operated by paul@keymail.dev added task Slice A: plan14 root-reads census overlay (written, eight rows); ruling asked of the census owner on #280
- task updated: Slice B (branch replica-freshness-284): build to the revised design, copy review
- claude agent, operated by paul@keymail.dev added task Slice A: copy review of the verb output, alert sentences and console alert text (Paul)
- task updated: Slice A: box metric and alert for a restart-looping unit
- task updated: Moved to slice B (branch replica-freshness-284): alert for replica freshness
- claude agent, operated by paul@keymail.dev pushed 3716414f5 slice A only: take the replica-freshness work back out (#284)
- claude agent, operated by paul@keymail.dev pushed b110af296 WIP: slice B as built on the withdrawn design, plus the console's alert detail (#284)
- claude agent, operated by paul@keymail.dev replied to #284 slice B design: replica freshness for both replicators, and where it shows
- claude agent, operated by paul@keymail.dev pushed f75dde2d6 replica freshness: the alert rule, the replicator's level-at, the status shape (#284)
- claude agent, operated by paul@keymail.dev opened #284 slice B design: replica freshness for both replicators, and where it shows
- task updated: Slice A: size carlos-litestream MemoryMax from the database count
- task updated: Slice A: back off the litestream restart loop and stop it flooding the journal
- claude agent, operated by paul@keymail.dev pushed 516ffbd25 alerts: a restarts rule for a platform unit that cannot stay up (#284)
- claude agent, operated by paul@keymail.dev pushed 940d59ea9 litestream: size the memory cap from the database count, back off restarts (#284)
- claude agent, operated by paul@keymail.dev added task Slice B (branch replica-freshness-284): build to the revised design, copy review