On 29/02/2012 9:10 p.m., Yucong Sun (叶雨飞) wrote:
PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND
1783 proxy 20 0 569m 529m 4132 S 1 26.4 274:53.22 squid
but the system still thinks it is using 529m memory
Kind of matches the "Total size:" mallinfo() was reporting.
To quote the FAQ:
"most operating systems do not allow processes to shrink in size.
When a process gives up memory by calling /free/, the total process size
does not shrink. So the process size really represents the maximum size
your Squid process has reached. "
I'm still on 3.1.15, I will see how it goes for 3.1.19
Squid Cache: Version 3.1.15
configure options: '--disable-maintainer-mode'
'--disable-dependency-tracking' '--disable-silent-rules'
'--enable-inline' '--enable-async-io=8'
'--enable-storeio=ufs,aufs,diskd' '--enable-removal-policies=lru,heap'
'--enable-cache-digests' '--enable-underscores'
'--enable-follow-x-forwarded-for' '--disable-translation'
'--with-filedescriptors=65536' '--with-default-user=proxy'
'--enable-ssl' '--enable-ltdl-convenience'
--with-squid=/tmp/squid-3.1.15/squid-3.1.15
On Wed, Feb 29, 2012 at 12:06 AM, Amos Jeffries wrote:
On 29/02/2012 6:50 p.m., Yucong Sun (叶雨飞) wrote:
Memory usage for squid via mallinfo():
Total space in arena: 536788 KB
Ordinary blocks: 173203 KB 4895 blks
Small blocks: 0 KB 0 blks
Holding blocks: 1420 KB 3 blks
Free Small blocks: 0 KB
Free Ordinary blocks: 363584 KB
Total in use: 174623 KB 32%
Total free: 363585 KB 68%
Total size: 538208 KB
These are numbers provided by the operating system. Squid is using ~174 MB
now and under peak traffic load it used ~538 MB. The difference has already
been free'd.
Memory accounted for:
Total accounted: 40844 KB 8%
memPool accounted: 40843 KB 8%
memPool unaccounted: 497364 KB 92%
memPoolAlloc calls: 0
memPoolFree calls: 841528260
That indicates the extra 500 MB as being temporary objects for processing
client requests as they pass through Squid.
92% unaccounted is strange though. If you have a Squid older than 3.1.19
please try upgrading, it could be one of several memory problems which we
have fixed already.
I'm also aware of a patch which can be tried on top of 3.1.19 as a last
resort. It is untested and a bit risky in 3.1 series.
Amos
