Forum Discussion

mnagel's avatar
Icon for Professor rankProfessor
7 years ago

alert expiration for "No Data"

I am not quite sure how this would end up working out, but I know the current behavior is not really working out too well.  I am having some trouble articulating the issue to LM support and in general, but here goes... 

If an instance  goes into alert and then later is removed from the underlying device (in reality, not in LM), the instance datapoints all generate "No Data" and the threshold alert will never clear until the instance is removed from LM.  It seems to me that after a long enough time of "No Data" the alert should clear on its own, but I am not sure how long that ought to be.  Certainly true for months of no data, but probably less than months.

What I am sure about is that in this specific case, which was for NetApp LUN volume usage, none of the datapoints alarm when the LUN is removed and THAT seems wrong as I should darn well have a way to know if a LUN goes away unexpectedly.  IMO, there should be a canary datapoint for anything critical like that marked to alarm on no data.  If there are no such datapoints, I do not understand why DS developers would not just set instances to auto-delete.

My main goals here are twofold -- we should not have alerts forever for stuff that is gone, and we should know when stuff that matters is gone and not be bothered when it does not matter.  And when I say "we", I mean my clients, who get frustrated by nonsense alerts, as they should.