Add location awareness to path outages
Description
Acceptance / Success Criteria
Lucidchart Diagrams
Activity
Alejandro Galue June 19, 2017 at 3:36 PM
I can confirm that without Minions involved, the path-outage feature works on H20.
Here is what I did on my test lab:
1. Install H20
2. Make sure path-outages feature is enabled on poller-configuration.xml
3. Start OpenNMS
4. Enable Notifications
5. Add the test requisition to monitor my Cisco lab running on GNS3 and synchronize it (this requisition has the parent-children relations defined, and the location field is not set on any node).
6. Make sure the path outages page shows the information correctly
7. Suspend a parent node.
8. Make sure that a notification was sent only for the parent node (notifications for the children are suppressed).
9. Make sure that all the nodeDowns alarms are generated (for the parent and the children).
10. Make sure that all the pathOutage alarms are generated (for the children).
11. Bring the parent node back up.
12. Verify the alarm/notification workflow.
Alejandro Galue June 19, 2017 at 3:23 PM
I'll re-build the lab and report back soon.
Jesse White June 19, 2017 at 3:07 PM
does the same path outage setup work without the Minion involved i.e. when all nodes are in the Default location?
Alejandro Galue June 14, 2017 at 12:02 PM
Unfortunately, this is not working on 20.0.0.
No pathOutage events, no nodeDown events updated, and the notifications are being sent for all the nodes regardless the path-outage configuration.
Let me know what's required from my test environment to give it a try.
Seth Leger January 17, 2017 at 3:43 PM
Merged PR, marking as fixed.
commit 6247e8a3cd1b1005c07890469beb85e29f33e67d
When using path outages in OpenNMS, pollerd attempts to ping the critical path IP address, when a node is found to be down. The result is added to the nodeDown even so that notifications can be suppressed if the critical path is unavailable, helping prevent a storm of notifications.
With the introduction of polling devices at remote locations, the critical path IP addresses may not be reachable from the OpenNMS instance, and cannot be directly pinged.
Path outage support should be enhanced to support pinging the critical path IP addresses using the same location as the node being monitored.
Future enhancements could include making the location of the critical path configurable.