Re: squid under solaris 2.5.1 consume a lot of file descriptors?

According to Bill Petersen's mail...
>
> HOW do you tune the tcp_close_wait_interval?

I have tried,

ndd -set /dev/tcp tcp_close_wait_interval 5000 but it made no
difference. Those CLOSE_WAIT are still there. I have changed it back
to the default afterwards.

So K T, CUHK

> Bill
>
> SO Kwok Tsun wrote:
>
> > Hi,
> >
> > I have configured squid 1.1.20 on a Ultra Sparc with Solaris 2.5.1
> > recently. I notice that when the connection to outside is sluggish,
> > there will be a number of CLOSE_WAIT with child connecting its http
> > port:
> >
> > parent.8080 child1.4501 33580 0 8760 0 CLOSE_WAIT
> > parent.8080 child2.4428 17520 0 8760 0 CLOSE_WAIT
> > parent.8080 child2.4963 17520 0 8760 0 CLOSE_WAIT
> > ...
> >
> > * parent is the Ultra Sparc.
> >
> > So, the number of file descriptors used increased a lot during that
> > period. It rises up over 1000 even when there are a few established
> > connection only. However, I seldom notice that there are any CLOSE_WAIT
> > in a squid parent proxy server under Digital Unix.
> >
> > I would like to ask whether you have similar observation. If your proxy
> > server is busy enough, have you tuned anything such as
> > tcp_close_wait_interval of /dev/tcp? Its default is 240000ms (=4mins).
> > Thanks a lot for your opinion.
> >
> > So K T, CUHK
>
>
>
> --
> ------------------------------------------------------------------------
> Bill Petersen email: brp@dsccc.com
> UNIX System, Web, Firewall, and Email Admin voice: 972-519-4249
> Technical Services, DSC Communications Corp, Plano, Texas
> ------------------------------------------------------------------------
> Communication, its one of those things we take for granted...
> ------------------------------------------------------------------------
>
>
>
Received on Wed Jan 21 1998 - 18:50:31 MST