What's the best way to test the datastore failover to standby? Stopping datastore service on primary machine and stopping primary machine altogether doesn't failover. Why not?
The Data Store won't fail over by just stopping the primary Data Store service or shutting the machine down because the fail over process is a bit intensive. Examples of situations that should cause a failover are if the primary loses network connectivity, (unplugging the ethernet cord), or hardware failures, (the machine crashes).
I have arcgis datastore 10.5, I've tried to shutdown the primary datastore so that the standby promotes to primary but this did n't happened although the installation and configuration for both datastore machine was successful. Also when they are both on, they appears to be healthy. I can manually set the standby machine to primary through rest "makePrimary". If the primary is shutdown, then making standby as primary (using makePrimary) will not succeed and results in the following:
Could not connect to the ArcGIS Server on machine 'WEBGIS2.domain.COM'.The ArcGIS Server service on that machine may not be running or the machine may not be reachable at this time.
webgis2 here was the primary before shutdown.
Did I miss a failover option during setup?
Stopping the Data Store service doesn't cause a failover, as that's considered a deliberate action. To cause a failover, the machine needs to have a disk or network failure, (again, not simply stopping/shutting down the machine).
How can i emulate a network or disk failure then? i need a proof that i setup a valid ha environment for our clients
Simulating a disk failure may be hard, but if you're on a VM, you can shut down the VM through the VM management console instead of shutting down the actual machine. If it's a physical machine, unplugging the ethernet cord simulates a network outage.
Thank you Jonathan, you solved my problem
Old question and answer, But as this article came up for a user I'd like to update this answer.
The datastore can be made primary from the server admin endpoint. In server administrator (Not manager)
Machines > Whichever is the new machine
This is a way to fail over by software command.
Retrieving data ...