On 10/1/2012 7:21 AM, tuxg wrote:
Looks like we overlooked the cache mem part. It looks fine. Let me also
correct that load increased once and we were using SquidGuard that time.
Sorry for the confusion.
total used free shared buffers
cached
Mem: 3791 3068 723 0 640 2001
-/+ buffers/cache: 425 3365
Swap: 4000 0 4000
Total: 7791 3068 4723
Do suggest if you see any scope for improvement in memory/cache_dir/other
settings. We expect upto 50 simultaneous users using squid at any point of
time.
Thank you.
tuxg wrote
Hi,
On an Intel i3 system with 4GB RAM, we noticed that squid is using
up all the memory without any limit. Using the binary that came with
CentOS 5.8 Final, rather than compiling from source. Does it have any
memory leak issues? Enabling memory pools has made not much
difference. Disabled SquidGuard and not much load on the server either
due to squid or due to other processes. Load increases only when it
uses up almost all the memory, but it is not using swap space even
then. Caching DNS server set up and runs on the same system.
Here are related settings:
<SNIP>
Suggestions from me:
if you are using squidguard i'm for replacing it with external acl
helper with concurrency to make it more efficient and use DENY_INFO with
arguments to make it all work.
What lists are you using with squidguard?
I wrote something that suppose to be external acl for squidguard domains
black lists.
core I3 with 4GB ram should be more then you need.
and try to use squid 3.2.1 that is the last stable version.
5.8 has 3.1.X version which is the older one.
Regards,
Eliezer
