Re: Squid NOVM stability

In article <19971214093158.31858@pointer.teuto.de>,
Lars Marowsky-Bree <lmb@pointer.teuto.de> wrote:
>Our main proxy (squid 1.1.18, Linux 2.0.32, 128 MB ram, 7 GB disk, on
>average ~4500 hits/hour) is slowly consuming all available mem (it is set to
>32 MB ram but instead is already using 125 MB, go figure).
>
>I have been thinking about installing the NOVM version. Is this likely to
>solve the problem? Which other resources does squid NOVM consume?

Well, I'm running squid-novm-1.1.17 on our main proxy box, and it has been
running well:

$ date
Sun Dec 14 15:38:00 CET 1997
$ psgrep squid
    USER PID %CPU %MEM VSZ RSS TT STAT START TIME COMMAND
proxy 16922 5.8 75.8 98440 97024 ? S Nov 20 2034:15 squid -D -s -f /et

As you can see it has been up for 24 days now. This machine has 2x4GB
striped as disk cache, but I need to lower the high water mark a bit:

$ df
Filesystem 1024-blocks Used Available Capacity Mounted on
/dev/md0 8572561 8125390 2633 100% /var/spool/squid

>Would I need to patch Linux to provide more filehandles?

Well I upped it to 1024, but we don't really seem to need it:

File descriptor usage for squid:
        Maximum number of file descriptors: 1024
        Number of file descriptors in use: 59
        Largest file desc currently in use: 82
        Available number of file descriptors: 965
        Reserved number of file descriptors: 256

We do get about 5 connections per second though:

Connection information for squid:
        Number of TCP connections: 3309854
        Number of UDP connections: 7812359
        Connections per hour: 19290.2
        Select loop called: 82934334 times, 25.028 ms avg

Mike.

-- 
 Miquel van Smoorenburg |  Studying to be a technomage   <*>
    miquels@cistron.nl  | "May you live in interesting times"