linkedin / venice

Venice, Derived Data Platform for Planet-Scale Workloads.
https://venicedb.org
BSD 2-Clause "Simplified" License
487 stars 84 forks source link

[controller][server][dvc][vpj][producer][consumer] Fixed sys store init #1238

Closed FelixGV closed 2 weeks ago

FelixGV commented 3 weeks ago

This fixes in an issue in NetworkClient::leastLoadedNode which manifested as a flood of IllegalArgumentException when some Kafka partition became completely unavailable.

How was this PR tested?

It's about to be.

Does this PR introduce any user-facing changes?

FelixGV commented 3 weeks ago

Integ test 1000 failed multiple times... there might be a regression. Will need to investigate before this can be merged...

FelixGV commented 2 weeks ago

Integ test 1000 seems to be flaky independent of this change, but I still tried to mitigate the flakiness in my latest commit... it's not very clean, IMHO, but it apparently did the trick in the one run after my commit went in...

FelixGV commented 2 weeks ago

@sixpluszero I added some unrelated mitigations to a flaky integration test, but the main change is still the same, and checks pass now. Would you mind taking another look when you get a chance? Thanks a lot 🙏