Re: [squid-users] Re: Hot Objects suddenly decrease

From: Amos Jeffries <squid3_at_treenet.co.nz>
Date: Wed, 27 May 2009 11:06:02 +1200 (NZST)

> Hi,
>
> at the moment that the hot objetcs decrease, i found it at caches log:
>
> FATAL: xmalloc: Unable to allocate 4869 bytes!
>

Aha.

> Squid Cache (Version 2.6.STABLE20): Terminated abnormally.
> CPU Usage: 22116.090 seconds = 12932.083 user + 9184.007 sys
> Maximum Resident Size: 0 KB
> Page faults with physical i/o: 2
> Memory usage for squid via mallinfo():
> total space in arena: -29064 KB
> Ordinary blocks: -31607 KB 161200 blks
> Small blocks: 0 KB 0 blks
> Holding blocks: 13216 KB 6 blks
> Free Small blocks: 0 KB
> Free Ordinary blocks: 2542 KB
> Total in use: -18391 KB 116%
> Total free: 2542 KB -15%
> 2009/05/25 14:40:53| ctx: exit level 0
>
> Someone knows if it is already fixed in 2.7 version ?

Nothing to be solved.
When the OS runs out of memory or squid exceeds its available cap bad
things happen. What you need to do is figure out how to prevent any such
memory overrun.

By the look of all those negative counters, I think your Squid is 32-bit
limited and may be trying to allocate more memory than it can handle. Try
a newer version and see if the larger counters fix your issue.

One common mistake is not knowing the cache_mem is the _minimum_ amount of
RAM Squid uses (allocated as RAM-cache), with all the indexes and live
running stuff on top of that.

Or, not having enough RAM on the box (Squid will run with as low as 32MB,
but very slowely).

Amos

>
> Thanks!
>
>
> On Tue, May 26, 2009 at 1:00 AM, Amos Jeffries <squid3_at_treenet.co.nz>
> wrote:
>>> Hi,
>>> I found the same problem in squid 2.6.21, too.
>>> I guess that the problem would be caused by max object size in memory.
>>> The parameter is likely not to work in some situation.
>>> In the next month, if I have some time, I'm going to trace it.
>>>
>>> If somebody already fixed the problem, please let me know please.
>>
>> If fixed its very likely fixed in 2.7 or 3.1 and require an upgrade.
>> The weird negative counters issue your trace shows was fixed at some
>> point, but thats only cosmetic AFAIK.
>>
>> Looking forward to anything you can find on this.
>> Amos
>>
>>>
>>> THW
>>>
>>>
>>> -----Original Message-----
>>> From: Juca!!! [mailto:yurass_at_gmail.com]
>>> Sent: Tuesday, May 26, 2009 2:52 AM
>>> To: squid-users_at_squid-cache.org
>>> Subject: [squid-users] Re: Hot Objects suddenly decrease
>>>
>>> Hi guys,
>>>
>>> could someone help me?
>>>
>>> I think its a bug!
>>>
>>> Tks
>>>
>>> Best Regards
>>>
>>> On Thu, May 21, 2009 at 1:42 AM, Juca!!! <yurass_at_gmail.com> wrote:
>>>> Hi guys,
>>>>
>>>> could someone help me indentify this problem, i use
>>>> squid2.6-stable20, and suddenly my Hot Objetcs in Mem Cache
>>>> decreases , it was storing 160k objetcs, and suddenly it decreases for
>>>> 50k objetcs, I get theses info at mgr:
>>>>
>>>> Memory usage for squid via mallinfo():
>>>> Total space in arena: -1190304 KB
>>>> Ordinary blocks: -1382006 KB 674308 blks
>>>> Small blocks: 0 KB 0 blks
>>>> Holding blocks: 13932 KB 5 blks
>>>> Free Small blocks: 0 KB
>>>> Free Ordinary blocks: 191701 KB
>>>> Total in use: -1368074 KB 116%
>>>> Total free: 191701 KB -15%
>>>> Total size: -1176372 KB
>>>>
>>>> And i get theses msgs at log:
>>>>
>>>> 2009/05/19 20:46:30| squidaio_queue_request: WARNING - Disk I/O
>>>> overloading
>>>> 2009/05/19 20:54:19| squidaio_queue_request: WARNING - Disk I/O
>>>> overloading
>>>> 2009/05/19 20:54:19| squidaio_queue_request: Queue Length:
>>>> current=448, high=468, low=321, duration=20
>>>> 2009/05/19 20:54:19| ctx: exit level 0
>>>>
>>>> What could I do ? Is it a bug?
>>>>
>>>> Thanks any help
>>>>
>>>
>>>
>>
>>
>>
>
Received on Tue May 26 2009 - 23:06:16 MDT

This archive was generated by hypermail 2.2.0 : Wed May 27 2009 - 12:00:02 MDT