[haiku-depot-web] Re: Server down?
- From: Andrew Lindesay <apl@xxxxxxxxxxxxxx>
- To: haiku-depot-web@xxxxxxxxxxxxx
- Date: Mon, 09 Jan 2017 18:16:04 +1300
Hello Axel;
However, I would still assume that there is some other player involved
that causes those peaks, or is there not?
Last time it was a daemon that was known to leak memory on Suse. At
that time, you adjusted the run-script to cause it to restart
periodically so that it would no longer consume all of the host's
memory;
Am 27/10/2016 um 09:43 schrieb Axel Dörfler:
| However, given that we have an actual culprit, buying more memory
| doesn't sound like the preferred solution; I have changed the nscd
| configuration to restart every two hours [1].
This time, I did not capture diagnostics, but remember that there were a
large number of "git" related processes in the log around the time of
the failure. Wild and unsubstantiated guess; it may be that the host
was IO-constrained and so was handling inbound-requests slowly and those
processes piled-up and consumed too much memory. Next time it happens I
will try to get details if I can.
We also don't have any IPs left, and the resources on our server aren't
endless either -- so that would be a last resort option IMO.
OK.
Other related posts:
