Re: [squid-users] memory utilisation

From: Edward <edward@dont-contact.us>
Date: Sat, 21 Apr 2001 12:32:15 -0400

It was doing this before FROX.

Disdk is working better than async-io.

I config async and storeio together.

When I reboot linux, I have over 100MB free. After squid starts with over
9GB of cache, it quickly falls to 3MB free.

After awhile, i start to se the swap going up. It nevers goes pass 7MB.

----- Original Message -----
From: "Joe Cooper" <joe@swelltech.com>
To: "Joe Cooper" <joe@swelltech.com>
Cc: "Edward" <edward@cariaccess.com>; "squid" <squid-users@squid-cache.org>
Sent: Saturday, April 21, 2001 12:29 PM
Subject: Re: [squid-users] memory utilisation

> Ok...So I misread the top. It's 6.4MB of swap used. Probably not a big
> deal. But if Squid is among the processes having bits swapped out,
> you're still in for trouble. (But a 91MB Squid process size isn't all
> that big.)
>
> Joe Cooper wrote:
>
>
> > That's 64MB of swap used in your top output. Squid is taking 94MB,
> > which seems about normal for a Squid with 8GB of cachedir running diskd
> > (why diskd on Linux? it's a less effective alternative to asyncio--but
> > if memory is a concern, don't use either). Something else is using a
> > large plot of memory on the system as well, though it's not in your four
> > lines of top ouput. If your box is being used primarily for Squid (and
> > I recommend it, since you are running an ISP through it), then you
> > should remove all other memory eating programs. If Frox is what is
> > taking up the rest of the memory, then you should probably remove it or
> > add more memory. Running Squid on a box that is going into swap memory
> > is not effective, and will be slow (as you've reported your Squid is
> > several times).
>
> > Edward wrote:
> >
> >> Hi everyone!
> >>
> >> I have a minor problem here.
> >>
> >> I am running Squid 2.4 up to 20th of this month, Linux 2.2.17, 192MB,
> >> 2 9GB
> >> UW SCSI (part 4GB each) and the OS and swapfile on an IDE hard drive.
> >>
> >> Cache drive is running diskd
> >>
> >> After a day or so, my squid w/ cache_mem =8 goes to 64Mb of memory
usage.
> >>
> >> Here is the TOP command:
> >>
> >> 10:12am up 23:30, 1 user, load average: 0.19, 0.26, 0.21
> >> 73 processes: 61 sleeping, 3 running, 8 zombie, 1 stopped
> >> CPU states: 8.5% user, 16.8% system, 0.0% nice, 74.6% idle
> >> Mem: 191684K av, 188572K used, 3112K free, 24988K shrd,
39236K
> >> buff
> >> Swap: 40120K av, 6432K used, 33688K free
27384K
> >> cached
> >>
> >> PID USER PRI NI SIZE RSS SHARE STAT %CPU %MEM TIME COMMAND
> >> 992 squid 20 0 94196 91M 3188 R 24.4 49.1 520:35 squid
> >> 8321 root 2 0 1040 1040 816 R 0.7 0.5 0:01 top
> >> 994 squid 0 0 1328 1328 748 S 0.1 0.6 0:38 diskd
> >> 997 squid 0 0 1328 1328 748 S 0.1 0.6 0:14 diskd
> >>
> >> I am also using FROX for FTP caching.
> >>
> >> What can I do to reduce my memory usage since I am using the GNU
> >> Malloc with
> >> seem to work better than the dmalloc the comes with squid.?
> >>
> >> Thank you very much.
> >>
> >> Best regards,
> >>
> >> Edward Millington
> >> (Network Administrator & Senior Technical Support Technician)
> >> Cariaccess Communications Ltd.
> >> Carlisle House
> >> Hincks Street
> >> Bridgetown
> >> Barbados
> >> 1-246-430-7435
> >> Fax : 1-246-431-0170
> >> www.cariaccess.com
> >
> >
> > --
> > Joe Cooper <joe@swelltech.com>
> > Affordable Web Caching Proxy Appliances
> > http://www.swelltech.com
>
>
> --
>
> --
> Joe Cooper <joe@swelltech.com>
> Affordable Web Caching Proxy Appliances
> http://www.swelltech.com
>
>
Received on Sat Apr 21 2001 - 10:32:06 MDT

This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:59:25 MST