Currently alert policies tigger events if ANY of the conditions in the policies are met. Example: if error percentage is high (above threshold %) then send an alert. I’d like to send an alert if the error percentage is high AND the number of requests in a given time window is exceeded. At this time we receive a lot of false positives on the weekends because we might receive 4 requests in a minute and 1 of them results in an error. Due to web crawlers etc. we almost always have a baseline number of errors in our system. On the weekends, when traffic on our product is low, these errors result in a lot of false positives because it rises above our percentage thresholds.
I’m afraid this isn’t available at the moment but I’ve gone ahead and dropped a feature request in for you!
I have to agree with this request from cfarnham.
At our large online retail business we are used to having to “mute” some of our alert triggers (from other tools) during times of low business volume. This is because a percentage deviation (a RELATIVE measure) needs to be backed up by an ABSOLUTE value below which the alert will NOT fire.
This is to counter the effects of Variance when dealing with low volume of business transactions.
Way to easy for this to be random fluctuation (variance).
We would want there to be an additional alert check that considered the number of users being sampled before triggering the alert.
Thanks for the example! I’ll pop this is as an addition to the Feature Request
+1 for this feature request. if required We can provide plenty of specific examples of problems this will solve…
In the absence of this functionality It would nice if at the very least there was an ability to suppress alerts within specific time windows.
If anyone has a suggestions on a better way of doing this than scheduling an api call to disable/re-enable alerts please let me know.
Hey @Trickey - I see you also opened a ticket. Going to add your +1 on this and reply in-kind to your ticket.
This is actually very important for my teams as we currently connect monitoring to call outs, using new relic we would be facing calls at 1 am because someone had a error causing a 100% error rate
@brian.lewis I have let our product management team know of your interest in this feature.
Thanks for the continued support for this alerting feature!
Hi checking in a Year later to see if any progress has been made ?
Hey @thomas_dooley the Alerts team has been focussing on some other priorities but compound conditions are one of the things they’re hoping to bring to you soon. Keep an eye out in the coming months for new capabilities! I can’t promise you’ll get everything on your wish list but there are a lot of good things happening in Alerts.
Has any progress been made on preventing spurious alerts from firing during low-traffic conditions?
I was hopeful that NRQL alerts would make this easier, but it doesn’t look like I can alert on multiple values in NRQL either. For instance, what I’d want to do is fire an alert if the Apdex is below 0.5 AND the number of requests is higher than 5 per minute, but I can’t see a way to do that.
Of course, I might just be underestimating NRQL alerts! That’s my hope, anyway.
We have no update on compound alert conditions yet. However, I will happily submit a feature request for you – this is a great idea!
I also recommend that you start a thread in our Alerts Feature Ideas section (https://discuss.newrelic.com/c/feature-ideas/feature-ideas-alerts). That way, other community members can chime in with their use cases and even vote on this, which will increase visibility with the product development team.
This might be helpful. I found a way to create an NRLQ query that with generate a chart with multiple APDEX scores plotted. This can be put on a monitor for the command center. It can be done with the following NRLQ query:
SELECT apdex(duration, .1) as ‘APDEX’ from Transaction since 1 hour ago
WHERE appName = ‘Search UI’ facet name where name in (
) limit 100 timeseries
With this query, I can monitor 44 APDEX charts in one. If the individual APDEX scores are relatively spread out over the range from 0 ~ 1 this view can be quite useful. With a single glance, you can quickly identify offenders corresponding to key transactions within a monolithc Application.
Has there been any update on this at all? I can’t seem to see it for application based alerts?
To build a NRQL query with looking at multiple sets of data, like in the example above, looking at apdex thresholds for multiple different Transaction names.
I’ll get your +1 added for native support of multiple conditions.