Influxdb: [2.0] InfluxDB notification rule doesn't send Slack message on status change

[mildebrandt]

__Steps to reproduce:__

1. Create a check, slack notification endpoint, and notification rule using the "changes from" condition.

__Expected behavior:__
A slack notification to be sent when the state changes.

__Actual behavior:__
No notification happens.

__Environment info:__
Docker container of InfluxDB 2.0.0 beta 8

I don't know if it's only Slack, but that's what we're testing with. When the rule is changed to the "is equal to" condition, we get the slack message successfully. So the endpoint and the check work correctly.

When I look at the history of the notification rule, the table says the state changed to "crit" was seen and it says a notification was sent. But we don't get any notification in Slack.

[desa]

@mildebrandt thanks for opening this issue. We've got a fix in place that has been making its way into the code base. As a temporary work around, if you don't use changes from then everything should work as expected. Can you confirm that if you switch to using is equal to that everything behaves as expected?

[mildebrandt]

Thanks for the update, looking forward to that!

Yes, "is equal to" works....but of course sends multiple messages until it's resolved.

Another related thing we noticed, we have a check that's scheduled every 2 minutes and a notification that's scheduled every 10 minutes. We get 5 notifications every time the notification fires (I'm assuming for the 5 times the check ran). Does that sound right to you, or do we need another issue opened?

[desa]

@mildebrandt how many series are in the check? You can figure out how many series are there by running

from(bucket: "_monitoring")
  |> range(start: v.timeRangeStart, stop: v.timeRangeStop)
  |> filter(fn: (r) => r["_measurement"] == "statuses")
  |> filter(fn: (r) => r["_check_id"] == "<id of my check>")
  |> limit(n: 1)
  |> group()
  |> count()

[mildebrandt]

So, the check recently was changed to check every 1 minute...and the notification was changed to every 2 minutes. When I run your query for that check, it's 2.

[ajkerrigan]

I'm happy to see this issue and response :). For the record I hit the same behavior for HTTP notifications (and assumed I was doing something wrong). I'm guessing the fix that's in the works is independent of destination, just figured this was worth noting in case any other non-Slack folks come across this. Thanks!

[samhld]

@mildebrandt clarification question: when you get 5 (or 2, in the subsequent case) notifications, are the Checks evaluating to a state that _should be_ triggering notifications? Or is there a notification for each of the Check windows in the Rule interval no matter what the Check evaluates to?

[mildebrandt]

I think I understand what you're asking. Let's see if this helps:

Check every 2 minutes. Notification on Crit every 10 minutes.

t0 -> OK (Check sees OK, notification sees OK, doesn't do anything)
t6 -> OK (Check sees OK)
t7 -> Crit
t8 -> Crit (Check sees Crit)
t9 -> Crit
t10 -> Crit (Check sees Crit, Notification sends two slack messages for both Crit checks)

So, I'd get two slack notifications for both checks that saw the Crit status. I would like just one.

I don't know what happens if the status goes back to OK at t10, whether I'd get one slack notification or no slack notifications. Personally, I'd expect no slack notifications at this point....but I could see someone else wanting to know if the check peaked into the Crit level within the notification time window.

Hopefully that helps.

[samhld]

Ok, thanks @mildebrandt !

[mildebrandt]

I found out the answer to what happens if t10 is OK, the Crit notifications get sent. Here are the checks:

And the notifications:

I was sent one notification at 23:10 and six notifications at 23:20.

[desa]

I believe that we have recently fixed this issue in cloud 2 and should have the fix available in the next influxdb oss release. Specifically, the state changes functionality should now work.

[mildebrandt]

That's great, I'll watch for the next release. Thanks!

[ajkerrigan]

Just writing in to note that my state change notifications did indeed start working after that last release. Yay and thank you!

[mildebrandt]

Sorry, I forgot to come back. This works for me now as well. Thanks!