[squid-users] Squid, Kerberos and FireFox (Was: Re: leaking memory in squid 3.4.8 and 3.4.7.)

[Victor Sudakov]

Carlos Defoe wrote:
> I had this kind of 100% CPU problem with auth helpers when upgrading
> to squid 3.4. I use negotiate_wrapper, kerberos, ntlm and basic auth.
> Then I had to fall back to 3.3 and it is production until now, with
> some troubles with broken clients, but with normal cpu usage most of
> the time.
> Can you try with squid 3.3 and see if that version don't have the same problem?

Carlos,

After downgrading to squid33-3.3.13_2 I have not received a single
complaint either from the users or from my monitoring system, and
squid's WCPU is about maybe 20% during office hours, briefly rising to
70% or so.

Thanks for the advice. 

Something nasty seems to have happened between 3.3 and 3.4.

Eliezer Croitoru wrote:
> Victor two things:
> - - you should consider that one cpu is not enough to operate a loaded
> proxy with auth.
> - - squid and the helpers needs CPU...
> - - 200 users can require a lot of resources which for you seems like
> nothing but from computing point of view it's something else.
> - - about the number of negotiate children.. it depends if you use
> concurrency or not and also in the number of requests per second.

Eliezer,

How come that the same load is tolerated by squid33 and not tolerated
by squid34? 

I'll see however, perhaps there will be a different usage pattern on
Monday and squid33 will choke too. I'll report.


-- 
Victor Sudakov,  VAS4-RIPE, VAS47-RIPN
sip:sudakov at sibptus.tomsk.ru