Re: [squid-users] xcalloc Fatal (squid 3.1.9)

On 10/03/11 18:54, V�ctor Jos� Hern�ndez G�mez wrote:
>
>>> I have found a message such as:
>>>
>>> FATAL: xcalloc: Unable to allocate 1 blocks of 536870912 bytes!
>>> Looks like a strange big block.
>>>
>>> My squid versi�n is 3.1.9. Any suggestion?
>>
>> Probably bug http://bugs.squid-cache.org/show_bug.cgi?id=3113
>>
>> Please try an upgrade to 3.1.11 to resolve that and a few other smaller
>> leaks.
>
> thank you, Amos, I have already upgraded squid to 3.1.11.
>
> In 3.1.9 we also got messages warning in that way :
> squidaio_queue_request: WARNING - Queue congestion (they were no very
> frequent)
>
> So I raised aufs_threads from 100 to 192 in compilation time. With very
> few activity I have received again the same warning. �Any idea or
> recommendation? �May be 192 threads is a very high value?

It is a sign that your disk speed is slower than your traffic speed.
Squid is getting ahead of the disks IO.

When that warning is displayed Squid will expand by double its queue
space. And when traffic slows down again the disks will catch up. So
nothing to worry much about if it occurs at startup or reconfigure, or
under traffic spikes.
   The time to worry is when it occurs repeatedly under normal loads. In
these cases you will need to spread that traffic load over several Squid
or the disk IO load over more disks and/or faster disks.

Amos

-- 
Please be using
   Current Stable Squid 2.7.STABLE9 or 3.1.11
   Beta testers wanted for 3.2.0.5