On Wed, Jul 18, 2007 at 09:05:23AM +0200, Dirk Kastens wrote:
is there a fix for the problem, meanwhile? We have the same situation, here. A filesystem on one of our mail servers ran full, but the admins didn't get an alert from hobbit.
James and I managed to track down the cause of his problems, and it turned out to be a configuration problem - specifically, the way the DURATION parameter in hobbit-alerts.cfg works.
James had used the DURATION setting to limit the number of alerts sent, by using "DURATION<30" to only send alerts for 30 minutes. Also, there was one group of people receiving yellow alerts, and another group receiving red alerts. His setup was like this:
HOST=foo SERVICE=disk
MAIL adam at foo.com COLOR=yellow DURATION<30
MAIL brian at foo.com COLOR=red DURATION<30
If the "disk" status went yellow at 6PM and red at 7PM, then brian at foo.com didn't receive any notification.
That's because the DURATION value is counted from the start of the event, which begins when the status goes yellow. So by 7 PM the event has a duration of 60 minutes, which is above the 30-minute threshold - so the red alert was suppressed.
The server is listed in our bb-hosts file on the "mail" page with an IP of "0.0.0.0". I found out that only the hosts with a real IP address will get an alert.
The IP in bb-hosts has nothing to do with alerts.
page linux 123.456.78.9 my.mail.host # ... page mail 0.0.0.0 my.mail.host # noconn page redhat 0.0.0.0 my.mail.host # noconn prefer
and I define alerts for all three pages, the alert only works for page "linux" and page "redhat". The host on page "mail" is being ignored.
This is a different problem. Each host has a "primary" page - only one! It's the first page that defines it (that would be "linux"), except if you use the "prefer" keyword then it is of course the page that has the preferred definition of the host ("redhat", in your example). If you're unsure of what page Hobbit uses as the primary page, then check it on the "info" status page.
Regards, Henrik