nv-l

AW: Return codes in "In-Line Action" Rul

1998-12-10 02:47:00
Subject: AW: Return codes in "In-Line Action" Rul
From: Schiffinger Ralph 0714 <Ralph.Schiffinger AT ERSTEBANK DOT AT>
To: nv-l AT lists.tivoli DOT com
Date: Thu, 10 Dec 1998 08:47:00 +0100
Hi!
Why exactly ping it at all?
Why bother with rulesets?
Why put additional load on the network?
Down is down and that's that.
My approach is fine-tuning ovsnmp.conf
(Options/SNMP Configuration).
My ovsnmp.conf currently has 94 lines,
which of course IS work, but:
Reasonable timeout- and retry-count-settings
should do (and - at least in "my" network - really do).
Regards, Ralph.

 -----Ursprüngliche Nachricht-----
Von:    Boyles, Gary P [SMTP:gary.p.boyles AT INTEL DOT COM]
Gesendet am:    Mittwoch, 09. Dezember 1998 22:44
An:     NV-L AT UCSBVM.ucsb DOT edu; Schiffinger Ralph 0714
Betreff:        Re: Return codes in "In-Line Action" Ruleset nodes


Why exactly are you pinging it five times?  I'd think
once were enough  (if you got a response back).

If you're going to do a re-check  (let's say because
of network conjestion)... I'd set the count to 5,
and maybe the wait-time (ie. -i) to 10.  You'd obviously
have to change the ruleset to timeout at about 1 minute.

If you're doing this because a node is bouncing up & down,
then I'd use the "Reset on Match" rule, and have both
interface up & down  (I wouldn't use node up/down) plug
into it.  I'd also set the "reset-on-match" wait time to
above a minute.

Note:  If you use interface up/down, you'll probably be
       wanting to look at the 1st word in the 4th attribute,
       instead of the node-name (i.e. 2nd attribute).

My 2 cents.

Gary Boyles


 -----Original Message-----
From: Giuliotti, J. P. [mailto:JGiuliotti AT MFS DOT COM]
Sent: Wednesday, December 09, 1998 12:35 PM
To: NV-L AT UCSBVM.UCSB DOT EDU
Subject: Return codes in "In-Line Action" Ruleset nodes


Hello Again,

AIX 4.2 NV6000 4.1
When NV reports a Node Down event (NDWN_EV), I run it through a ruleset that
pings the node from an In-Line Action box with the following settings:
Command: ping -c 5 $NVATTR_2
                                Wait interval: 20 seconds
                                Command Exit Code Comparison: Not Equal to:
0

This tells me that if Netview doesn't see a node it will try to ping it 5
times and will give it 20 seconds to respond. If it sees a return code other
than zero (there is a problem), then it will forward it on to the next node
in the ruleset.

I am still getting false alarms. Am I misunderstanding the way an inline
action works? Is 20 seconds enough time for 5 responses? (I think so). What
if 2 out of the five pings are "time outs" due to network latency? Will that
trigger the return code to return other than a zero value even though the
other 3 pings were good?

Thanks in advance.


J.P. Giuliotti  - Massachusetts Financial Services
(617) 954-5684
JPG AT MFS DOT COM

<Prev in Thread] Current Thread [Next in Thread>
  • AW: Return codes in "In-Line Action" Rul, Schiffinger Ralph 0714 <=