[squid-users] How to analyse squid memory usage

[Dan Charlesworth]

No worries—thanks for following up on it!

That’s very interesting, about the concurrent requests, because the “normal” report does around 80% more requests per day than the “leaky” one — a few hundred thousand vs a couple of million.

Does this CLOSE_WAIT sockets issue have a bug being tracked or anything like that? I’ve probably overlooked the discussion on the list.

> On 1 Jun 2016, at 10:26 PM, Amos Jeffries <squid3 at treenet.co.nz> wrote:
> 
> Hi Dan,
> sorry RL getting in the way these weeks.
> 
> Two things stand out for me.
> 
> Its a bit odd that exteral ACL entries shodul be so high. But your
> "normal" report has more allocated than the "leaky" report. So thats
> just a sign that your external ACLs are not working very efficiently
> (results being fairly unique, so the lookup cache not being much use there).
> 
> In the "leaky" report there are 10K concurrent requests still active.
> Normal report shows only 1K requests. So up to 10x the state data
> storeage is needed by that proxy.
> 
> 
> I'm a little suspicious you might be seeing another symptom of the issue
> behind what others have been reporting as too many CLOSE_WAIT sockets
> staying open with Squid not doing anything for them.
> 
> Amos