On 07/03/2017 05:44 AM, Miroslav Suchý wrote:
Dne 27.6.2017 v 16:17 Kevin Fenzi napsal(a):
- retrace01.qa.fedoraproject.org is almost constantly alerting on swap
being full. Not sure what to do about this, but perhaps we could add more swap or somehow limit it to use only memory for normal jobs?
Few months ago I set postgresql to use more agressive caching. So that is main culprint for consuming so much memory. I can easily lower it by few percent. But... I see right now that there is 16GB swap and 8 GB is free. And total available memory is 16 GB. Because 8GB free swap and 8GB are kernel buffers/cache. So when you see those errors and what are the exact numbers in those alerts?
retrace01.qa.fedoraproject.org
Looks like it alerted just a few min ago:
Swap
Notifications for this service have been disabled CRITICAL 06-27-2017 14:15:24 0d 0h 11m 8s 3/3 SWAP CRITICAL - 7% free (1011 MB out of 16383 MB)
Swap-Is-Low
Notifications for this service have been disabled CRITICAL 06-27-2017 14:15:03 0d 0h 11m 29s 4/4 SWAP CRITICAL - 7% free (1002 MB out of 16383 MB)
OK. I lowered the DB cache settings. Please let me know if this happen again.
ok. It looks like it's good so far. ;)
The only thing left on our nagios report is retrace02 being 100% full on disk. ;)
kevin