Open another-rex opened 1 week ago
Manually bumped the memory to 8Gb from 4Gb, and will continue monitoring
This seems to have stopped OOM errors now, though there is a lot of 20 second queries now at 99%> . Theory is that this is caused by the paging limit correctly kicking in and preventing a timeout, and now that there is more memory these requests are not killed by OOM.
Tracking issue for the GRPC out of memory error. We likely need to bump the memory given the recent large OSV records we are now serving.