-
Notifications
You must be signed in to change notification settings - Fork 584
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Notifications are sent when reloading Icinga 2 even though they're deactivated via modified attributes #4696
Comments
Updated by mfriedrich on 2016-09-30 13:08:34 +00:00
Do you happen to have any debug logs from those reloads which could help identify and reproduce the issue you're having? |
Updated by pstiffel on 2016-09-30 15:13:09 +00:00 The problem is still happening, so I guess, I can switch on the debuglog and try to reproduce the error. |
Updated by pstiffel on 2016-10-07 11:31:41 +00:00 What is the preferred way to post the debug log? It is a production machine, so I don't want to include the debug.log here. It's too much work to anonymize the log... |
Updated by mfriedrich on 2016-11-18 15:32:56 +00:00 I'd prefer a specific host/service and notification example extracted from it (that works well with replacing sensitive names too). That flow is important to see how they pass or not pass the specified state/type filters, or if the notification is for example forced if it shouldn't be. Other details which are helpful - specific configuration objects from "icinga2 object list --type ..." for those objects (Host, Service, Notification, User). |
Updated by pstiffel on 2016-11-18 15:47:58 +00:00 Ok, I see what I can do. |
Updated by pstiffel on 2016-11-25 14:52:39 +00:00
Managed to declutter and anonymize some debug logfiles and the service definitions. Let me know if you need more or if I removed important lines... |
Updated by mfriedrich on 2016-12-07 21:41:33 +00:00 The configuration files all have "enable_notifications = true" set. So does the debug log tell - there are reminder notifications in place which will check against such parameters, in addition to state/type filters. Everything looks normal from this point of view. Can you tell me how you did disable notifications for these services? |
Updated by pstiffel on 2016-12-12 06:47:52 +00:00 The services failed and we turned the notifications off via the classic UI. |
Just a quick feedback: the same thing happens with V2.6 |
Might be a race condition between loading the state file and the modified attributes file. Can you verify that both HA instance have modified-attributes.conf set to the same values (enable_notifications = false). |
Could this be related to #4946 ? We also checked what you suggested: On both nodes, the modified-attributes.conf says that the notification is disabled. |
Hm, this log entry looks suspicious.
Can you try the following:
I would believe that issue is gone then. The issue is somehow that one node thinks that notifications are enabled, and enforces that over the cluster. I'm not sure yet how to reliably reproduce the issue. |
Hmm, this issue is not related to a dedicated host. |
+1 |
Same here, also with HA - seems to be a timing problem when reloading. The notifications are obviously sent before the files from /var/lib/icinga2/api or something else are loaded, this is where the downtimes and ACKs from the API/Icingaweb are stored afaik. |
There are two problems with are possibly related at the event of triggering re-notifications.
We're currently investigating on a possible race condition between activating all objects (including the NotificationComponent feature) while not having applied modifications and runtime state. One thing you can do to help nail down the issue:
Reproducing the issue
Test Config
Attached to this issue: TestsTest case 1Config with 3 hosts and 300 services each. Run once and disable notifications via API. Send a kill command to icinga2, don't do foreground + ctrl+C. Verify inside modified-attributes.conf that everything is disabled.
See that the log says its sending notifications.
Test Case 2Increase the host and service count. Modify Will continue with debugging and investigating on a fix. |
FYI, the current patch doesn't address the ::IsActive issue yet. |
Patch resolves the issue with modified attributes only. The issue could still happen with downtimes not activated on notification request. |
I'm moving the second part of the patch to a separate ticket (#5224) because we're not going to target 2.6.4 for this one just yet. |
Can someone please confirm if the fix for the issue "modified attributes are not applied" will be released in 2.6.4 or 2.7.0 and also is there any bugfix version being released for this issue? |
At this point in time this ticket's milestone is set to 2.6.4, i.e. it'll be released as part of the next minor release. There's no ETA for that release yet. |
This issue has been migrated from Redmine: https://dev.icinga.com/issues/12817
Created by pstiffel on 2016-09-28 08:40:48 +00:00
Assignee: pstiffel
Status: Feedback
Target Version: (none)
Last Update: 2016-12-12 06:47:52 +00:00 (in Redmine)
Unfortunately, even after updating to 2.5.4, Icinga2 still sends out notifications for services which have notifications disabled, when icinga2 is reloaded due to configuration changes.
And it is not deterministic which of the services get notified.
Example.
At the moment, we have 3 services which are in a critical state and which have notifications disabled. After a reload (we added some hosts and services), mail, sms and ticketsystem-notifications were send out for only one of the services.
The next reload due to configuration changes produced mail and sms notifications for all 3 services, but a third notification (email to a ticketsystem) was only notified for 2 services.
We have an HA-cluster-setup running on two nodes (Debian Jessie with packages from debmon.org)
Attachments
The text was updated successfully, but these errors were encountered: