Hi Dave, quick question...
We are building quite a big node list in freenats these days and a lot of the nodes we need quite regular testing on, ie they are set to 1 minute intervals, this is for grpahing/reporting purposes, some of these are critical nodes and we need to be alerted immediately if they flounder, others although we want the frequent monitoring on, we can live with a few minor blips in service...so...
I was wondering if it is possible to have a threshold limit
ie: Node tests every minute, and rather than alerting on every fail, could it be set to only alert if fails repeatedly, say 3 or 4 times on the trot?
Thanks
Matt
Alert thresholds
Re: Alert thresholds
Hi Matt,
Not exactly possible but bodgeable in the current system in a couple of different ways:
Per-Test Retries
Tests support custom retry and timeout values. For example a web test could be set to retry 3 times with a 60 second timeout. This would be probably better than, for example, the node being tested only every three minutes with 0 retries (try once) and a 60s timeout in terms of not reporting minor connectivity glitches but the failed tests aren't recorded anywhere if the last one passes (the last test result is recorded).
Event Based Logic
Using the event handlers it would be pretty simple to turn on alerting on a node when it had failed more than X times. On the next check if it failed an alert would be generated (previous failures being recorded but not alerted upon). It would then turn the alerting flag on once a node had passed more than once.
If this would suit you let me know and I'll knock up a custom handler for you. You'd need to edit a file to add/remove nodes from your "retry" list but could do that via the file manager in the interface.
As for implementing it as a fully supported feature I'll let you know if I think of a way it can be accomplished without too much code-bodging.
Cheers,
Dave.
Not exactly possible but bodgeable in the current system in a couple of different ways:
Per-Test Retries
Tests support custom retry and timeout values. For example a web test could be set to retry 3 times with a 60 second timeout. This would be probably better than, for example, the node being tested only every three minutes with 0 retries (try once) and a 60s timeout in terms of not reporting minor connectivity glitches but the failed tests aren't recorded anywhere if the last one passes (the last test result is recorded).
Event Based Logic
Using the event handlers it would be pretty simple to turn on alerting on a node when it had failed more than X times. On the next check if it failed an alert would be generated (previous failures being recorded but not alerted upon). It would then turn the alerting flag on once a node had passed more than once.
If this would suit you let me know and I'll knock up a custom handler for you. You'd need to edit a file to add/remove nodes from your "retry" list but could do that via the file manager in the interface.
As for implementing it as a fully supported feature I'll let you know if I think of a way it can be accomplished without too much code-bodging.
Cheers,
Dave.
Re: Alert thresholds
thanks Dave, i'll have a fiddle with the retry counts first of all and see how I get on with that...
Matt
Matt