Cluster Controller failover problem

Jacky Lv1Posted Aug-27-2025 17:51

I built a cluster using 2 nodes without witness node and did some HA tests. Let's say A is the Cluster Controller and B is the other node.

When I unplug the power of B, those VMs on B can be successfully recovered on A in a few seconds.

However, when I unplug the power of A, those VMs on A failed to be recovered on B, and the Cluster Controller didn't failover automatically. I had to manually set the B as Cluster Controller, those VMs finally came back.

So my questions are:
1. Does Cluster Controller have failover feature?
2. If yes, is lack of witness node the reason why my Cluster Controller didn't failover?

Thank you.

By solving this question, you may help 792 user(s).

Posting a reply earns you 2 coins. An accepted reply earns you 20 coins and another 10 coins for replying within 10 minutes. (Expired) What is Coin?

Enter your mobile phone number and company name for better service. Go

Newbie337989 Lv2Posted Aug-28-2025 10:31
  
Based on you concern

2 nodes only (A + B), no witness.
- A = current Cluster Controller (CC).
- When B fails → A survives, so quorum is intact, VMs restart fine.
- When A fails → CC is lost, and with no witness, cluster arbitration fails → B cannot automatically promote itself to CC. - That’s why you had to manually set B as CC.

1. Does Cluster Controller have failover feature?

- I believe yes, in a properly configured cluster, the Cluster Controller (CC) can failover to another node. The CC role is not fixed to one host; another node can take over if the original CC fails.

2. Is the lack of witness node the reason why CC didn’t failover?

- Exactly. In a 2-node only cluster without a witness, there is no quorum mechanism.
- Quorum = “majority vote” to decide which node should be CC.
- In a 2-node setup, if 1 node goes down, the surviving node cannot be sure if it’s the only survivor, or if it’s split-brain (network partition).
- Without a Witness Node (the 3rd vote), the surviving node won’t automatically promote itself to CC.
- That’s why your VMs didn’t failover until you manually set B as CC.

Try this

- 2 Nodes + Witness Node (VM or physical) > Required for HA and CC automatic failover.
- The Witness Node acts as an arbitrator (tie-breaker).
- In your scenario, if A (CC) fails:
  - Witness + B form majority (2 out of 3).
  - B is promoted to CC automatically.
  - VMs from A restart on B without manual intervention.

I Can Help:

Change

Board Leaders

NyxZale...

Weekly Sharers

Newbie5...

Weekly Questioners