[hobbit] Large number of TIME_WAITs

Buchan Milne bgmilne at staff.telkomsa.net
Tue Jan 29 17:18:24 CET 2008


On Tuesday 29 January 2008 17:53:59 Gary Baluha wrote:
> >  Wouldn't it make sense if you had a Hobbit client running on the same
> > server as the Hobbit server?
>
> But that many TIME_WAITs?

On a server receiving monitoring information and monitoring itself a total of 
~ 150 hosts, I have ~ 500 hobbit ports in TIME_WAIT, and about 200 htto ports 
in TIME_WAIT.

More answers in-line (ug, thread re-assembly ...).

>
> > =G=
> >
> >
> >
> > *From:* Gary Baluha [mailto:gumby3203 at gmail.com]
> > *Sent:* Tuesday, January 29, 2008 10:36 AM
> > *To:* hobbit at hswn.dk
> > *Subject:* Re: [hobbit] Large number of TIME_WAITs
> >
> >
> >
> > Red Hat ES 4 Update 5.

same on mine.

> > That's what I normally would think, but it doesn't make any sense that
> > the source AND destination are both localhost or the local IP.  It's as
> > though the Hobbit server is trying to communicate with itself using
> > network sockets.  It would make more sense if either the source or
> > destination was a client IP, but that isn't the case.

It is the case, bbtest-net connects to hobbitd. The client on the server also 
does. Any other server-side extensions may. bbgen may also (haven't checked, 
but I assume ...). The cgi's will probably also ...

> >
> > On Jan 29, 2008 10:15 AM, Hubbard, Greg L <greg.hubbard at eds.com> wrote:
> >
> > What OS?  Some operating systems allow network connections to hang around
> > a bit after disconnection.  On Solaris you can set a parameter to reduce
> > the timeout period.

On Linux, it is net.ipv4.tcp_fin_timeout, e.g.:

net.ipv4.tcp_fin_timeout = 30
in /etc/sysctl.conf, defaults to 60s.

Regards,
Buchan



More information about the Xymon mailing list