Re: MemPools rewrite

On Mon, Nov 13, 2000, Andres Kroonmaa wrote:
>
> After a polymix stresstest, totally, absolutely idle squid is
> holding:
>
> Poolname inUse
> Store Mem Buffer 1077
> mem_node 1078
>
> MemObject 51
> HttpReply 51
> HttpHeaderEntry 465
> HttpHdrCc 49
>
> First it looked like some requests got stuck somewhere in squid. All
> the tests were finished 2 days ago, no client sockets nor server
> sockets are open.
> Can someone familiar with the code explain whats the possible cause
> for that? Seems that some stuff is sitting in storemem cache. Why
> are some objects in state "swapout: xxx bytes queued"?

I think thats just a leftover from the swapout. SWAPOUT shouldn't be
set otherwise.

> Why I'm asking is because I'm seeing some considerable freespace
> fragmentation (160M in mallinfo Total free) right now, and I'm
> wondering if it would be a right thing to add timebased timeout to
> inmem object cache, after which unreferenced objects are dropped
> from ram. Should allow squid to compact its memory usage some time
> after a spike of high activity...

Hrm. My storage manager fixes will definitely fix this. You *COULD* set
maximum_object_size_in_memory to say, 1k .. :-)

Adrian

-- 
Adrian Chadd			"God: Damn! I left pot everywhere!
<adrian@creative.net.au>	  Now I'll have to create Republicans!"
				    - Bill Hicks