On 10/26/2017 03:58 PM, Vieri wrote: > I should've done this before, but I just took a second look at > cache.log and found the following message repeated a few times before > the 100% CPU issue: > "Your cache is running out of filedescriptors" That warning matches the dangerously low number of available descriptor reported by your mgr:info: > File descriptor usage for squid: > Maximum number of file descriptors: 1024 > Largest file desc currently in use: 913 > Number of file desc currently in use: 791 > Files queued for open: 0 > Available number of file descriptors: 233 > Reserved number of file descriptors: 100 > If this is the cause then I'm even more worried now because: > > open files (-n) 16384 > # cat /etc/security/limits.d/01_squid.conf > * hard nofile 65535 > * soft nofile 16384 Your Squid does not know about the above configured limits. It thinks the limit is 1024. > I don't get it. There aren't that many active file descriptors (less than 1000). ~900 descriptors is "a lot" when the limit is 1024. > Also, on low traffic (late evening) with no CPU or mem issues at all, > mgr:filedescriptors shows that there are about 2500 active file > descriptors. Please note that "2500 active descriptors" contradicts the 1024 limit reported in mgr:info in your earlier email (quoted by me above). Squid also reports the limit at startup. Here is an example (not from your logs): > 2017/10/26 15:48:04.744| With 65535 file descriptors available There are many email threads and probably some wiki documents that discuss increasing the number of file descriptors available to your Squid and numerous related caveats. Cheers, Alex. _______________________________________________ squid-users mailing list squid-users@xxxxxxxxxxxxxxxxxxxxx http://lists.squid-cache.org/listinfo/squid-users
