Repository navigation
Redis scheduler store: RESP3 cursor pages, the retention default, an unreadable row skipped, a wait for the new index - #2803
Merged
Conversation
|
The latest updates on your projects. Learn more about Vercel for GitHub.
|
amankrx
force-pushed
the
pr/redis-record-hygiene
branch
from
September 28, 2026 11:24
373843f to
72a4246
Compare
…arses (cherry picked from commit 1b85fc8c97a3346e4e792795f707bcdd4ef1d275)
… records expire (cherry picked from commit 183a533c2e565f462f866b8f160c0e55a3e1b19a)
(cherry picked from commit ceb68a3aebda279802384c65e604961675b9e107)
(cherry picked from commit 955d1fbcf6212d78862d76da77b3cbadef6920f1)
amankrx
force-pushed
the
pr/redis-record-hygiene
branch
from
September 28, 2026 18:35
72a4246 to
1af0384
Compare
MarcusSorealheis
left a comment
Member
There was a problem hiding this comment.
LGTM thank you for fixing
This branch was successfully deployed
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
What and why
Four small fixes to the Redis-backed scheduler store, each found on a test cluster. A cursor page after the first came back in the RESP3 shape the first page already parses, and failed. The Redis backend never applied the 60 s retention default, so completed records lived until evicted. The queue listing ended at a record it could not read instead of skipping it. And a freshly created search index was read before its initial scan finished, so the first listing after a restart missed rows.
How was this verified?
Pinned command sequences in
redis_store_test.rscover the RESP3 page shape and theFT.INFOwait afterFT.CREATE, and fail on the old code. The listing is tested with a fault-injecting database: a row the store cannot decode and a record that cannot be read are each left out and the rest listed, in both the queued and the all-operations listings, the sweep retires the rest, and a read failing because the store is unreachable is reported rather than skipped; those tests fail without the change. The retention default is one helper shared by both backends, with a test. On a test cluster, completed records now carry a TTL, a restart's first pass read the whole queue, and a corrupted record no longer ended the listing. The runbook gains the one-offEXPIREpass for records written before this.Risk
Low. Completed records expire after 60 s on Redis as they always did in memory; a deployment relying on them staying longer should set
retain_completed_for_s. The index wait adds oneFT.INFOround trip per index creation.AI assistance
An agent drafted the change and I reviewed every line.
This change is