Re: Delays and timeouts when telnetting to random open ports?
- From: nick.seccombe@xxxxxxxxx
- Date: 18 Feb 2006 06:00:30 -0800
Hello,
I think you need to look at this at a lower level.
I would check the behaviour is as you expect - run a snoop and check
that you are really not getting a TCP DR to your TCP SYN request which
is what I believe is the network underlying response when you get a
connection refused.
snoop -d qfe0 port 1234
Compare good with bad.
If you are getting no response, then try tracing system calls using
truss on the listener - compare good with bad bahaviour. Look for Err
messages, or log files being written to that you may not have been
aware of - you may find a full queue or the likes, a resourcing issue,
or a name service (dns, nis, ldap) timeout.
basic truss to follow children:
truss -f -p <pid of listener>
or for a log file:
truss -f -o /tmp/logfile -p <pid of listener>
or for really verbose output:
truss -f -wall -o /tmp/logfile -p <pid of listener>
You can also add timestamps to truss (-t a, or -t r I think).
Cheers,
Nick
mkirsch1@xxxxxxxxxxxxxxxx wrote:
I've got a real strange one going on here....
You know what happens when you telnet to an open port on a system,
right? It supposed to come back instantly with "connection refused."
Right?
We've got one odd system that cycles between that "normal" behavior,
and no response at all. In between are several levels of delays.
Again, it seems to be cyclical in nature: For a few seconds, it will
work normally. Then for the next few, the delays will get longer and
longer. Finally, for several minutes, the connection will simply time
out. Lather. Rinse. Repeat. It doesn't matter what port or what
interface you use, including the loopback.
What brought this on was a DBA trying to start up a custom Oracle
listener on port 1526. He's having the same type of problem testing the
startup and shutdown of the listener. If he catches it when things are
working, the listener is okay. However, if he catches it on the down
side of the cycle, the listener times out trying to grab the port.
There's another system that's virtually IDENTICAL to this one, both in
hardware and software, and it does not exhibit this same phenomenon.
showrev -p on both hosts is the same. /etc/resolv.conf is the same.
/etc/nsswitch.conf is the same.
.
- References:
- Delays and timeouts when telnetting to random open ports?
- From: mkirsch1
- Delays and timeouts when telnetting to random open ports?
- Prev by Date: Re: How to configure preinstalled Solaris 10
- Next by Date: Re: Is it possible to frig VxVM to get it to boot off a disassociated boot mirror?
- Previous by thread: Delays and timeouts when telnetting to random open ports?
- Next by thread: Prom configuring from running OS
- Index(es):
Relevant Pages
|