Always On failure testing (three nodes)

Test environment:

Cluster three nodes are normal

Witness that arbitration is normal:

Current number of votes, number of votes is odd: total number of votes is 3

Get-Cluster I ft name,DynamicQuorum,witnessDynamicweight

1, 1 node failure, Always On is normal, witness arbitration vote is 1, the total number of votes is 3

2, 2 node failure, Always On is normal

3. Witness arbitration failure and 2-node failure, Always On is normal

Summary: since the maximum number of Always On automatic failover is 2, DB03 cannot achieve automatic failover. When DB01 or DB02 is stopped at the same time, the number of votes will become DB03. In this case, DB02 failure will not affect the cluster and Always On service, and DB03 failure cluster and Always On will not work. It is clear whether the vote is random.

According to the test situation, you can put DB01, DB02 and witness on one server and DB03 on one server, so that the failure of any server does not affect business usage.