Add support for failover nagios deployment
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
Nagios Charm |
Won't Fix
|
Wishlist
|
Unassigned |
Bug Description
As Nagios is currently not cluster-aware, it is possible to deploy multiple nagios units which will both monitor and alert on the status of the environment through the notifications/
I would like to see an option to detect whether multiple units of nagios are deployed and to have the Juju elected-leader have the external (pagerduty) notifications enabled and have them disabled on the non-leader unit with some way for the units to monitor and alert against each other not able to reach the notification APIs and taking over alerts for the other. This is essentially to reduce double-incident alerting when running two nagios monitors.
Another consideration is to have nagios non-leader sitting idle with nagios not running, and on leader-elected, starting nagios if it's the new leader. This methodology would reduce the overhead of hitting all of the monitors on nrpe endpoints twice.
Third option may be to tie into hacluster and configure it to manage the active nagios instance and VIP for the nagios web endpoint.
Changed in charm-nagios: | |
importance: | Undecided → Wishlist |
Changed in charm-nagios: | |
status: | New → Triaged |
The third option makes much more sense to me. I don't think we should tie a critical piece of software like Nagios to Juju leadership, my feeling is that it doesn't offer enough guarantees. If we decide to go that route, I'd strongly suggest confirming with the Juju devs that it's a good idea.