Page 1 of 4

Lookback period issue regression in 1.4

Posted: Wed Jan 20, 2016 7:00 am
by weveland
Sorry for the clickbaity style title. But It looks like the lookback period issue I reported with the previous release may have returned. After upgrading to 1.4 I'm getting alerts for checks with lengthy lookback periods. I can view the alert in the dashboard and visually see the results. But the return is 0 events, hence the alarm.

--
Wayne

Re: Lookback period issue regression in 1.4

Posted: Wed Jan 20, 2016 7:03 am
by weveland
Ok deactivating and reactivating the alert got it working properly. So my initial assessment may have been incorrect. Any ideas what could have happened here?

Re: Lookback period issue regression in 1.4

Posted: Wed Jan 20, 2016 11:32 am
by rkennedy
Something may have changed with the configuration, are you seeing any alerts anymore or is it functioning as expected?

Re: Lookback period issue regression in 1.4

Posted: Wed Jan 20, 2016 11:35 am
by weveland
It seems to be fine at the moment.

Re: Lookback period issue regression in 1.4

Posted: Wed Jan 20, 2016 12:29 pm
by jolson
Any ideas what could have happened here?
Not sure - is it possible that your single alert has been misbehaving since before the fix was put in place? The following fix was implemented:
Fixed alert run end time slight offset on slow systems

The above bug was caused to to an inconsistency in the time that alerts were scheduled to run, and when they actually ran. There is a potential that the alert could miss some time, which may result in a missed alert here and there - credit to @Jklre for pointing this out.

After the alert subsystem was upgraded, something must have happened to cause your alerts to regress as they did - have you noticed any sort of inconsistency since making this post?

Re: Lookback period issue regression in 1.4

Posted: Wed Jan 20, 2016 12:32 pm
by weveland
Not so far. Before disabling and re-enabling the alert I did try manually running it a few times, so maybe it was modified in the config files. Then when I deactivated and re-activated it, it was pulled from the database and put down correctly.

That's my line of thought.

Re: Lookback period issue regression in 1.4

Posted: Wed Jan 20, 2016 6:08 pm
by hsmith
Let's monitor it for a couple of days to see if it comes back.

Re: Lookback period issue regression in 1.4

Posted: Thu Jan 21, 2016 1:11 pm
by weveland
Unfortunately same issue again this morning. Alert fired at a slightly different time this morning.

Yesterday: 6:31 AM
Had to deactivate and reactivate alarm to clear and return OK.

Today: 6:54 AM
Just had to re-run check manually and alarm returned OK.

Re: Lookback period issue regression in 1.4

Posted: Thu Jan 21, 2016 5:43 pm
by hsmith
How long are your check intervals/lookback set to? I want to test this on my end.

Re: Lookback period issue regression in 1.4

Posted: Fri Jan 22, 2016 10:34 am
by weveland
Checks run every 5 minutes. Lookback period is 5 hours.

As a side note. I didn't get an alarm this morning