On proxy40 the mirrorlist-server process is killed regularly by the OOM killer.
proxy40 seems to be one of the few system which only has 4GB of memory. Many other proxies have at least 8GB.
Each mirrorlist-server process currently needs around 1.1GB of memory and with all the other things running on proxy40 and without swap this seems to result in the OOM killer regularly ending one of the mirrorlist-server processes.
Memory of this system should probably be increased to 8GB.
I agree they should have 8G at least...
that said, it does all work if you reboot... after some amount of uptime, memory becomes fragmented and then you can't keep it running (it just OOM's after trying to start).
Metadata Update from @kevin: - Issue priority set to: Waiting on Assignee (was: Needs Review) - Issue tagged with: low-gain, low-trouble, ops
Metadata Update from @kevin: - Issue assigned to mobrien
@mobrien you were going to look at re-deploying those aws proxies with more memory?
This has finally been redeployed on a bigger ec2 instance
Metadata Update from @mobrien: - Issue close_status updated to: Fixed - Issue status updated to: Closed (was: Open)
Login to comment on this ticket.