Scheduler scaling - Githubissues

brietaylor commented 4 years ago

Quick gut-based fixes for the scheduler scaling problem.

Added the index we've been using to the database initialization
Stop querying "done" blocks from the monitor. We have to perform a sequential scan of the "done" blocks to produce the metric and it's possible that's killing our cache.

The latter change will remove "done" blocks from Grafana, which will break the speed meter. I do have a plan in place to get those metrics back.

ababaian commented 4 years ago

Does this 'clear' or remove 'done' blocks after merge_done state is reached?

brietaylor commented 4 years ago

No it doesn't. I think that would be slower. This just stops accessing those blocks via the metrics page, which I suspect might be forcing them into memory. Just a gut feel though, I'll need metrics to validate that hypothesis.

El vie., 12 de jun. de 2020, 7:17 p.m., Artem Babaian < notifications@github.com> escribió:

Merged #160 https://github.com/ababaian/serratus/pull/160 into master.

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/ababaian/serratus/pull/160#event-3440364681, or unsubscribe https://github.com/notifications/unsubscribe-auth/ABMJRWJ6UMPQCXQDWUUCXI3RWLOTBANCNFSM4N4U76HQ .

ababaian / serratus

Scheduler scaling #160