investigation on queues during sync
Closed this issue · 2 comments
We had a situation where
- CL is at head
- EL is syncing
- looks like a full pending pool queue
I'd like to get a better understanding of why, but I believe it's probably with the rationalisation that if the EL comes online we have the votes to track head and attest... It did on this occasion use 90% of cpu and the pending pool sat at roughly 30k constantly....
i did a quick check but only let it run on my node for a couple of minutes.. the queue did fill but we didn't get a lot of load as a result, so i think we're probably ok...
I think the load was potentially the EL sync, but lets setup a scenario and let it run more than a finality period and just make sure queue management across epoch is ok..
after the EL finished syncing above, the pool went back to normal, so I think we're ok, the load was from EL sync.