[Date Prev]   [Date Next] [Thread Prev]   [Thread Next] [Date Index]   [Thread Index]

 

     Re: [snips-users] Notifier vs sniplogd question

Thanks for the note. It has validated that I'm on the right track. Now, after have slept on the issue, I revisited the confg files. I finally noticed the -1 age for a notification. That was the last part I was missing. How did snipslogd send STDIN to notifier and send info to the right person.

So this is what I have as my default notifier line.

default chingmd:email:0 chingmd_pager:email:60 curtis:email:240 chingmd:email:-1

So:

If I understand things right.

When a system hits a critical state, I get email immediately, after it's at critical for 60 minutes I get paged, and curtis gets email at 4 hours. (all without hourly reminders)

chingmd:email:-1 only applies to the snipslogd daemon and piped commands.

Based on these entries:

*       Critical        |/tool/adclocal_home/snips-1.2b2/bin/notifier.pl^-
*       Warning         |/tool/adclocal_home/snips-1.2b2/bin/notifier.pl^-

When a monitor goes from info to Warning, it send me an email (based on the configured snipsperld and notifier.pl scripts). When it hits Error email is sent to me based on that it left Warning. And finally sends an email when it hits Critical.


So they work together, but piped data to notifier looks for a -1 age for STDIN.


WOOHOO!

Mark




At 09:16 AM 11/25/2002 +0100, Joerg Mertin wrote:
Hi Mark,

almost right. I use snips 1.1 - but I noticed the following to apply.
The snipslogd notifies everything that has something to do with the event you configured to trigger the execution of external script.


This means - the snipslogd is going to notify: info <-> critical, warning <-> critical, error <-> critical etc. It will notify events in both direction.

The snipslogd starts the notifier command using the "-" sign as argument, telling the notifier to handle STDINPUT only - the notifier script beeing started by cronjob does a timely preconfigured job and handles only the events that happened for a prespecified time, b.e. 1 hour.

Regarding on what the snipslog daemon does - I think, anyone correct me if I'm wrong, write the logs/* files, and if enabld the rrd-file data.

So - to make a long story short:
1. Notifier started by cron is notifying everything for a prespecified amount of
time.
2. Notifier started by snipslogd handles only device-lines that are passed to it
through snipslogd, enabling the user to have a more fine grained capability
of monitoring special devicenames, e.g. company-critical customer b.e.


The only drawback I find using snips the way it comes out of he box is that it has to heavily be adapted to easily monitor about 2500 devices... Especialy the updates-file stuff should be rewritten - as it has a big problem in old snips aka. 1.0 and 1.1 (don't know higher version). The following should be done for both version: got the updates informations written to a database (done that).
Implement a locking mechanism for updates-writing (1.0) plus for genweb.cgi (in 1.1) to not generate web-pages during a monitor-reload - as it erases some entries while the log-file synchronization takes place.


Cheers

Joerg

Mark Ching wrote:
Hi Folks,
I've recently installed Snips 1.2beta2 (yesterday) I've configured a few monitors and have been trying to figure out how the notifier and snipslogd work together for notification of downed services
Could someone ensure my sanity and let me know that what I'm seeing/thinking is right?
So far this is what I understand.
Notifier.pl is the script that sends an email or page, but only when an error hits the critical level. Then sends every hour if configured.
It doesn't send a notification when the service/machien/etc comes back.
snipslogd on the other hand more like a syslogd. It captures output on port 5354 and writes to log files configured in snipslogd.confg. And can be configured for piped commands.


Here is where I have gotten myself completely confused. Do the Notifier cronjob and syslogd piped command work mutually exclusive of each other?
So if you just want to know when things change state, then you should only use snipslogd?
Mark




--
Joerg Mertin                                 PSINet Europe
Business Process Automation Group Manager    Rue Fritz-Courvoisier 103
Phone: +41-32-967 52 00                      2300 La Chaux-de-Fonds
E-mail: mertinj at europe psi.com               Switzerland
        *** check the BPA-Pages: http://bpa.etc.psi.com ***


Zyrion Traverse Network Monitoring & Network Management Software