Re: [squid-users] squid becomes slow (when cache_dir is full)

On 06/07/11 21:30, Sergiu Tatar wrote:
>> Hello friends.... i have a problem with my squid server ( conf proc: Intel(R) Celeron(R) E3200 @ 2.40GHz, 6 GB ram dual-channel 800 Mhz, hdd S-ATA II 500 GB Cache 32 Mb ) , when cache_dir is full squid becomes slow.... i have 60 Mbps bandwith and i can't use only 10-15 Mbps with squid up. Linux kernel is: 2.6.33.7-desktop-2mnb #1 SMP Mon Sep 20 18:19:20 UTC 2010 x86_64 x86_64 x86_64 GNU/Linux
>> Squid Configure Options:
>> Squid Cache: Version 3.1.12.1-20110503
>> configure options: '--enable-cache-digests' '--disable-icmp' '--disable-delay-pools' '--enable-storeio=ufs,aufs' '--disable-ident-lookups' '--prefix=/squid2' '--enable-removal-policies=heap,lru' '--with-filedescriptors=16384' '--disable-snmp' '--enable-linux-netfilter' '--enable-follow-x-forwarded-for' '--disable-translation' '--with-large-files' --enable-ltdl-convenience
>
>> Squid config:
>> cache_dir aufs /mnt/squid 92160 64 512
>> cache_mem 2048 MB
>> maximum_object_size_in_memory 96 KB
>> maximum_object_size 24 MB
>> fqdncache_size 4096
>> ipcache_size 4096
>> ipcache_low 95
>> ipcache_high 97
>
>> Squid info:
>> Squid Object Cache: Version 3.1.12.1-20110503
>> Start Time: Wed, 29 Jun 2011 03:56:36 GMT
>> Current Time: Wed, 06 Jul 2011 09:07:51 GMT
>> Connection information for squid:
>> Number of clients accessing cache: 0
>> Number of HTTP requests received: 7677294
>> Number of ICP messages received: 0
>> Number of ICP messages sent: 0
>> Number of queued ICP replies: 0
>> Number of HTCP messages received: 0
>> Number of HTCP messages sent: 0
>> Request failure ratio: 0.00
>> Average HTTP requests per minute since start: 738.8

Only 12 req/sec? That does not compute with 60Mbps.

Your avg. object size indicates an avg around 276 Kbps. Weird, or most
of your requests are not cacheable and will be stressing the disks.

>> Average ICP messages per minute since start: 0.0
>> Select loop called: 886726996 times, 0.703 ms avg
>> Cache information for squid:
>> Hits as % of all requests: 5min: 36.1%, 60min: 40.9%
>> Hits as % of bytes sent: 5min: 6.4%, 60min: 8.8%
>> Memory hits as % of hit requests: 5min: 41.2%, 60min: 20.1%
>> Disk hits as % of hit requests: 5min: 28.5%, 60min: 33.4%
>> Storage Swap size: 90597584 KB
>> Storage Swap capacity: 96.0% used, 4.0% free

96% of the cache full. That is 1% over the "high water" and 6% over the
"low water" default settings.
  Assuming the cache_swap_high and cache_swap_low are still at those
defaults your Squid will be a bit busy erasing ~6GB worth of ~23KB files
from the disk. Could take a while, especially if the new arriving
traffic is creating files faster than the erasures. The fact it got much
above 90% is a sign that the traffic arriving _is_ faster than erase.

>> Storage Mem size: 2076808 KB
>> Storage Mem capacity: 100.0% used, 0.0% free
>> Mean Object Size: 23.09 KB
>> Requests given to unlinkd: 0
>> Median Service Times (seconds) 5 min 60 min:
>> HTTP Requests (All): 0.04277 0.04047
>> Cache Misses: 0.10281 0.09736
>> Cache Hits: 0.00000 0.00000
>> Near Hits: 0.04519 0.01955
>> Not-Modified Replies: 0.00000 0.00000
>> DNS Lookups: 0.03696 0.03696
>> ICP Queries: 0.00000 0.00000
>> Resource usage for squid:
>> UP Time: 623475.324 seconds
>> CPU Time: 24039.848 seconds
>> CPU Usage: 3.86%
>> CPU Usage, 5 minute avg: 6.34%
>> CPU Usage, 60 minute avg: 4.71%
>> Process Data Segment Size via sbrk(): 3264060 KB
>> Maximum Resident Size: 13229792 KB
>> Page faults with physical i/o: 402
<snip>

Amos

-- 
Please be using
   Current Stable Squid 2.7.STABLE9 or 3.1.14
   Beta testers wanted for 3.2.0.9