Arseny Sher

I took a quick look and didn't saw anything particularly suspicious in logs initially -- looked like just pageserver was too slow to chew full segment in one minute of...

@aome510 thanks for the investigation. > One way to fix this is to modify `WalreceiverState::applicable_connection_candidates` That would leave the connection open even if there is nothing to do (currently it...

> Is there anyone actively working on implementing this? I'm not too familiar with the safekeeper code, but can try if needed. @aome510 Yes, please. In case of questions feel...

> As far as I understand it is the same problem which I have fixed in https://github.com/neondatabase/postgres/pull/187 It looks like all your 3 variants allow to write (disable backpressure) while...

@yeputons Could you please try again on current main? After https://github.com/neondatabase/postgres/pull/187 merge it should be gone.

Superseded by [26746](https://github.com/neondatabase/neon/pull/2228)

Got it again https://github.com/neondatabase/neon/runs/7723174729?check_suite_focus=true

Hm. The first link https://app.circleci.com/pipelines/github/zenithdb/zenith/1303/workflows/d5d95541-50e7-41e3-8a14-5ce9e7cacddb/jobs/7372 points to 'pg_regress tests debug' green job where safekeepers are not involved, and in the second https://app.circleci.com/pipelines/github/zenithdb/zenith/1746/workflows/999057a9-bf5a-4d3e-8613-584716a2bf6e/jobs/11086 `test_race_conditions` failed, but I don't see `Cannot assign...

Haven't seen this for a while.

Haven't seen this for a while.