I have an alerts file where hobbit sends out an alert to a ever-widening group of people based on DURATION... the longer the un-ack'd duration, the more excited it gets.
The problem comes in when something goes YELLOW for a while and then goes RED. All the time it spent at YELLOW is included in the DURATION even through I have a COLOR=red explicitly given in the config file.
Is there a way to specify the duration as the time something has been RED as opposed to NOT-GREEN?
Thanks, Jon
It's a good question, and I'd like to know the answer too.
Does this alert correspond to a red conn, or some other test? Because I knew that yellow conn does not correspond to a recovery, but IMHO a yellow in any other test *should* correspond to a recovery (compared to a red). (Of course, there may be some tests some people have where it should not, so maybe it should be configurable somewhere... Maybe a RECOVERY= tag in the alerts file to specify what colour should correspond to a recovery? The existing tag is called RECOVERED.)
SebA
From: Jon Boede [mailto:jon at shadowsoft.com] Sent: 02 September 2008 23:25 To: hobbit at hswn.dk Subject: [hobbit] DURATION in alerts
I have an alerts file where hobbit sends out an alert to a ever-widening group of people based on DURATION... the longer the un-ack'd duration, the more excited it gets.
The problem comes in when something goes YELLOW for a while and then goes RED. All the time it spent at YELLOW is included in the DURATION even through I have a COLOR=red explicitly given in the config file.
Is there a way to specify the duration as the time something has been RED as opposed to NOT-GREEN?
Thanks, Jon
In <48BDBD4F.8020807 at shadowsoft.com> jon at shadowsoft.com (Jon Boede) writes:
I have an alerts file where hobbit sends out an alert to a ever-widening group of people based on DURATION... the longer the un-ack'd duration, the more excited it gets.
The problem comes in when something goes YELLOW for a while and then goes RED. All the time it spent at YELLOW is included in the DURATION even through I have a COLOR=red explicitly given in the config file.
Is there a way to specify the duration as the time something has been RED as opposed to NOT-GREEN?
Unfortunately not.
It's tricky, because a lot of people would also argue that if a status has been red for 10 hours, dips to yellow for 5 minutes and then goes back to red, then those few minutes of yellow should not be enough to reset the duration counter.
Of course, this only applies for a status that can go yellow (e.g. cpu- or disk-utilisation) - network tests usually don't have this issue.
Henrik
participants (3)
-
henrik@hswn.dk
-
jon@shadowsoft.com
-
spa@syntec.co.uk