2.2STABLE5+hno has /no/ memory leaks in the core parts of Squid (and in
any subsystem that I've used). We have some very heavily loaded sites
that have uptimes measurable in months running that version on Linux.
The Squid process will grow for as long as the contents of the cache
object store are growing. This could take a couple of weeks to reach
'full' if you're site isn't very heavily loaded. If this is the case,
and memory is a concern, you can lower the size of your cache_dir to
something that makes you happier wrt memory usage.
30GB of cache object store will pretty much fill your 512MB. I usually
only run about 24GB of cache on a 512MB machine--and the process size of
Squid runs up to over 200MB (but stabilizes when the cache is full).
2.4 probably has a few problems lurking in the dustier corners...but
seems pretty stable. But if your reason for switching is non-existent
memory leaks in 2.2STABLE5+hno, then you'd gain nothing by switching.
Greg wrote:
> I have being fighting a constant memory leak problem.
>
>
>
> i have tried dmalloc, and i am now at the moment trying gnumalloc, and i
> have tried the suggestions in the squid faq.
>
>
>
> does squid 2.4 have less or virtually no memory leaks when compared to
> squid 2.2.STABLE5 with henriks patches?
>
>
>
> when i do ps -aux. the memory % line on squid keeps going up, every day,
> its a pII 333, 512 of ram, and 30 gig scsi
>
> with L1 72, and L2 256.
--
Joe Cooper <joe@swelltech.com>
Affordable Web Caching Proxy Appliances
http://www.swelltech.com
Received on Wed May 09 2001 - 00:06:40 MDT
This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:59:53 MST