Re: Hanged INFORMIX connections
Hi Rafael.
You can modify TCP_KEEPALIVE using ndd command.
i.e)
How to get default TCP_KEEPALIVE_INTERVAL
[tsutoo@bygzam ~]% ndd -get /dev/tcp tcp_keepalive_interval
Change TCP_KEEPALIVE_INTERVAL
# ndd -set /dev/tcp tcp_keepalive_interval XXXX(millisecond)
Caution!!(My experience)
If you reboot your machine, tcp_keepalive_interval set to default (2HR)
You may put the following script in /etc/rc3.d(In my case)
% cat /etc/rc3.d/S81keepalive
#!/sbin/sh
/usr/sbin/ndd -set /dev/tcp tcp_keepalive_interval XXXX
--
Tsutomu Ogiwara from Tokyo Japan.
ICQ#:168106592
>From: rafapadilla@yahoo.com (Rafael Padilla)
>Reply-To: rafapadilla@yahoo.com (Rafael Padilla)
>To: informix-list@iiug.org
>Subject: Re: Hanged INFORMIX connections
>Date: 12 Apr 2005 08:31:03 -0700
>
>Martin Fuerderer <MARTINFU@de.ibm.com> wrote in message
>news:<1113298761.6d7d83a54b3619fff618e4b08ccc280e@teranews>...
> > Hi,
> >
> > right, TCP-KeepAlive should do that (according to its configuration,
> > which by default is not 300 seconds but 2 hours).
> >
> > Alternatively you can put the plug of the network cable back in
> > and everything should be fine as if never anything happened, as
> > long as the client application process is still there that is. This
> > is a "great feature" of TCP/IP, will work with normal telnet/rlogin
> > etc. sessions as well and will get the security paranoids amongst
> > us up on the highest palm tree in sight ... ;-)
> >
> > Regards,
> > Martin
> > --
> > Martin Fuerderer
> > IBM Informix Development Munich, Germany
> > Information Management
> >
> > owner-informix-list@iiug.org wrote on 11.04.2005 19:42:11:
> > > Rafael Padilla wrote:
> > > > Hi all,
> > > >
> > > > I realized that client connections to INFORMIX remained open if You
> > > > accidentally remove network cable; it's just to give an example.
>Other
> > > > kind of similar event did same thing.
> > > >
> > > > How can you monitor a connection to Informix with this problem?
> > > > When I reviewed user threads with onstat -u or onstat -g ses, They
> > > > seems to be in normal conditions, but they were lost for conditions
>as
> > > > I described above.
> > > >
> > > > Any tip?
> > > >
> > > > I got a SUN Box and IDS 9.40 FC3
> > > >
> > > > Thanks in advance!
> > >
> > > I would have thought that the network settings for TCP_KEEPALIVE would
> > > kick in after a while (300 seconds and then the retry period and
>number
> > > of times I think is the default); then they would be "cleanly tidied
>up"
> >
> > > because the connection would get a signal.
> >
> > sending to informix-list
>
>Thanks for your attention to this...
>
>Could you give an advice about TCP_KEEPALIVE ?
>is it a SOLARIS parameter?
>we use CLIENT/SERVER applications.
>
>regards
sending to informix-list