Unleash / unleash / 39224624618

88%

master: 91%

push

github

feat: per-instance bucket offset for user sync job

I want to increase the randomness of requests and only one job/endpoint
remains a problem regardless of what I did.
It seems that `singleInstance` is the problem here. 




I'm not 100% sure what `singleInstance` is doing or why. It seems like
something to pick the instance to run the job for each `clientId`, but
the way it picks it makes it so that most instances for all clients make
requests at the same time.

So this is about adding some randomness to that process and trying to
offset that a bit.

We could also increase the time of user syncs 🤔 

> ## About the changes
> All instances' buckets for `sync-users-addon` start at the same time,
so they tend to sync with auth-app right after each boundary.
> 
> - `JobService.singleInstance` / `JobStore.acquireBucket` take an
optional `bucketOffsetMs` (default 0)
> - `user-sync` derives the offset from a hash of `clientId`: pods of
one instance share buckets, while different instances' buckets are
spread out
> - Test covers the offset end-to-end through `JobService`
> 
> 🤖 Generated with [Claude Code](https://claude.com/claude-code)
>

4056 of 5165 branches covered (78.53%)

16485 of 18703 relevant lines covered (88.14%)

767.46 hits per line