Hi Dojo,
We configured a failover setup following the best practices described in the DataMiner Help [Link]:
We tested the following scenario:
Active and offline agents lose connectivity with a network switch that is used to configure the inverted heartbeats. Acquisition and Corporate networks are still available from active and offline agent.
When the network cable connecting to the switch is pulled out from the active and offline DMA, we noticed the following:
- In the Alarm Console, a notice will appear, indicating that the heartbeat path is failing
- The active agent goes offline
- The offline agent stays offline
After a couple of minutes, the active agent becomes online again.
Checking the FAQ related to failover setups [Link], we could not find this use case. Please could you let us know if the events described above are expected?
DMA version: 10.1.2.0-9866
Hey Miguel,
Scenario as far as I understood: only inverted heartbeats are failing (because of the switch not longer being reachable), and they are failing on both machines, other heartbeats are still successful.
It is expected for the active agent to go offline (because of the inverted heartbeat failure)
It is also expected that the offline agent stays offline (because of the inverted heartbeat)
The active agent going back online is somewhat unexpected, but can most likely be explained by the logic where the normal heartbeats can bring online the agent which was online most recently if it was detected that both agents are offline at the same time. ("Going online because partner XXXX if offline"). That's also what happens when restarting Failover agents (both start up as offline initially). I actually expect the agent to start toggling between online and offline in this case (the inverted heartbeat will bring the agent offline again later on).
In any case, the exact flow for the actions can be observed in the SLFailover.txt logfiles of both agents (in earlier versions: SLNet.txt)
A case could be made against bringing that agent back online. It can be argued that this action should not be taken while the inverted heartbeats are failing.
Hi Wouter,
Thanks for the detailed explanation.
Indeed, the scenario that you are describing is what I described in my initial question.
I understand that this is expected behavior. As long as both DMAs are not able to reach the network switch (using inverted heartbeats), both DMAs will stay offline.
I also understand that it is not expected to see a DMA (of the failover pair) become online after the connection with network switch is restored. Does it mean that, in an expected scenario, we will need to set manually an agent to become online?
I will create a task reporting the unexpected scenario.