Re: [hobbit] possible alerting bug in RC2?

[Henrik Stoerner <henrik (at) hswn.dk>]

On Mon, Feb 14, 2005 at 01:28:28PM -0800, Bruce Lysik wrote:

> So I installed RC2 this morning.  Later on, I noticed an alert email
>  for a monitor going into yellow.  I had disabled this previously
>  with --alertcolors=red,purple in hobbitlaunch.cfg.

[config from another mail]

>HOST=<snip list of about 100 hosts>
>        SCRIPT /opt/bb/server/ext/email bruce_mail FORMAT=SCRIPT REPEAT=30 DURATION>6 RECOVERED

> And here's the alert I just received:
> 
> im68:cpu yellow [-1]
> yellow Mon Feb 14 13:13:56 PST 2005 up: 208 day(s), 1 users, 115 procs, load=529

The alert you show here looks like a recovery-notice (the "-1" I
assume is the acknowledgment cookie, and this value indicates that
there is no active alert).

If you look in the ~/data/ack/notifications.log file for these
notifications, you can tell if it's an alert message or a recovery
message by the number of columns in the file. E.g. in my log I have

Wed Feb 16 13:08:43 2005 www.sslug.dk.smtp (130.228.2.150) henrik (at) hswn.dk 1108555723 725
Wed Feb 16 13:09:44 2005 www.sslug.dk.smtp (130.228.2.150)henrik (at) hswn.dk 1108555784 725 61

The first one is the alert message, the second is the recovery
message. The recovery has an extra field "61", which is the duration
of the event (in seconds).


Could you check the following in hobbitlaunch.cfg:

* The "hobbitd" command has "--alertcolors=red,purple --okcolors=green"
* The "hobbitd_alert" command has "--alertcolors=red,purple"

This setup should give you alerts when a status is red (or purple),
and recovery notices only when they go green (after being red or
purple).


Regards,
Henrik