By “base number”, I’m referring to the resources used used when there is minimal activity and no connections?
I’m running into timeouts, and I think it’s from file descriptor exhaustion (or CPU issues). I have a cron job running showing the number of open file descriptors:
Before Couchbase or Sync Gateway wake up, total file descriptors (from ‘sudo lsof’): 1435
8074 total 3307 beam.smp 1995 memcached 552 moxi 157 nginx 130 node 145 PM2 96 sync_gate
Do those numbers seem correct? High? Low? That’s on boot-up with no other activity. Anytime my total # file descriptors gets near 8192, the timeouts start popping up. And I have set my soft/hard /etc/security/limits.conf to 50000
From these numbers, I don’t understand the need of setting the ulimit > 10200 though, as that’s a per user number I thought?
Here is the cron job I’m using to get those numbers:
#!/bin/bash lsof > /home/sj/tmp echo "`date` `wc -l < /home/sj/tmp;` total `cat /home/sj/tmp | cut -d' ' -f1 | sort | uniq -c | grep -E 'beam.smp|memcached|moxi|nginx|node|PM2|sync_gate' | tr -d '\n'`" >> /home/sj/lsof-count.log
I’ve also noticed that my Couchbase Server uses 100% CPU for a couple of seconds every 2 minutes (like clockwork) - my page faults also go up to about 25k for that time too. What is happening that would cause this?