-----Original Message-----
From: Buchan Milne [mailto:user-9b139aff4dec@xymon.invalid]
Sent: Wednesday, January 25, 2012 9:02 AM
To: xymon at xymon.com
Cc: Root, Paul
Subject: Re: [Xymon] Can I monitor how many connections are in
TIME_WAIT for a specific port
On Tuesday, 24 January 2012 18:13:41 Root, Paul wrote:
Hi,
We are monitoring a particular port that we are having issues
with.
8022, it's a proxy port for HP NA.
Anyway, I have an expect script that goes in and tests the
functionality of the port. But when it starts to go bad, this script
get
stuck in TIME_WAIT, along with the users connecting to the port.
So, can I look at the port data before I try connecting, and
if
there are a bunch of TIME_WAIT connections, just skip the test
entirely?
Why don't you just (in hobbit-clients.cfg or analysis.cfg) use
something like
this
PORT LOCAL=%([.:]389) STATE=TIME_WAIT MIN=0 MAX=750 COL=yellow
PORT LOCAL=%([.:]389) STATE=TIME_WAIT TRACK=ldap-wait MIN=0
MAX=1500
COL=red
(example taken directly from a similar requirement for monitoring
highly
utilised LDAP servers with badly behaving clients - regex could
probably be
improved but works fine for my purposes)
I'm running the test from the xymon server, so I was thinking
of
pulling the data out of xymon directly. Would that by xymoncmd?
Why script around it when built-in features can detect and alert on the
error
condition (and provide graphs as well in case you want to correlate the
exact
number of connections in a specific state to other events)?
Regards,
Buchan