[squid-users] Squid 5.2 unstable in production mode

[Flashdown]

Hi David,

well I am curious, where did you set the max filedescriptors? Only in 
the OS configuration? If so, you also need to define it in the 
squid.conf as well -> 
http://www.squid-cache.org/Versions/v5/cfgman/max_filedescriptors.html

Regarding the memory leak, do you use an adaption service such as 
c-icap?
If so, what is the result of: ss -ant | grep CLOSE_WAIT | wc -l

May you should try to build Squid5 against Debian 11 to have the latest 
version of any dependencies needed to see if the memory leak is gone or 
not.

I run multiple Squid 5.2 servers on Debian 11 in production and do not 
have any issues.
---
Best regards,
Enrico Heine

Am 2021-11-11 20:08, schrieb David Touzeau:
> Hi
> 
> Just for information and i hope it will help.
> 
> We have installed Squid 5.1 and Squid 5.2 in production mode.
> It seems that after several days, the Squid become very unstable.
> We mention that when switching to 4.x we did not encounter these
> errors with the same configuration, same users, same network ( replace
> binaries and keep same configuration )
> 
> All production servers are installed in a virtual environment ( ESXI
> or Nutanix ) on Debian 10.x with about 4 to 8 vCPUs and 8GB of memory.
> and from 20 to 5000 users.
> 
> After severals tests we see that the number of users did not have
> impact with the stability.
> We encounter same errors on a 20 users proxy and the same way of a
> 5000 users proxy.
> 
> 1) Memory leak
> ---------------------------------
> This was encounter on computer that handle more than 10Gb of memory,
> squid eat more than 8Gb of memory.
> After eating all memory, squid is unable to load helpers and freeze
> listen ports.
> A restart service free the memory and fix the issue.
> 
> 2) Max filedescriptors issues:
> --------------------------------
> This is a strange behavior that Squid did not accept defined
> parameter:
> Example we set 65535 filedescriptors but squidclient mgr:info report
> 4096 and sometimes return back to 1024.
> 
> Several times squid report
> 
>     current master transaction: master15881
> 2021/11/11 17:10:09 kid1| WARNING! Your cache is running out of
> filedescriptors
>     listening port: MyPortNameID1
> 2021/11/11 17:10:29 kid1| WARNING! Your cache is running out of
> filedescriptors
>     listening port: MyPortNameID1
> 2021/11/11 17:10:51 kid1| WARNING! Your cache is running out of
> filedescriptors
>     listening port: MyPortNameID1
> 2021/11/11 17:11:56 kid1| TCP connection to 127.0.0.1/2320 failed
>     current master transaction: master15881
> 2021/11/11 17:13:02 kid1| WARNING! Your cache is running out of
> filedescriptors
>     listening port: MyPortNameID1
> 2021/11/11 17:13:19 kid1| WARNING! Your cache is running out of
> filedescriptors
> 
> But a mgr:info report:
> 
>         memPoolFree calls:    4295601
> File descriptor usage for squid:
>         Maximum number of file descriptors:   10048
>         Largest file desc currently in use:    262
>         Number of file desc currently in use:  135
>         Files queued for open:                   0
>         Available number of file descriptors: 9913
>         Reserved number of file descriptors:  9789
> 
> After these errors the listen port is freeze and nobody is able to
> surf.
> a just "squid -k reconfigure" fix the issue and the proxy return to
> normal mode for several minutes and back again to filedescriptors
> issues.
> 
> There is no relationship between filedescriptors issues and the number
> of clients.
> Sometimes the issue is discovered during the night when there is no
> user that using the proxy ( just some robots like windows update )
> 
> Is there something other we can investigate to help more stability of
> the 5.x branch ?
> Regards
> _______________________________________________
> squid-users mailing list
> squid-users at lists.squid-cache.org
> http://lists.squid-cache.org/listinfo/squid-users