Why ODD number of Voting Disks or Why 3 Voting Disks

Before we start discussing why do we need 3 voting disks or odd number of voting disk, why should know the Role of Voting Disk  

The Role of Voting Disks in Oracle RAC

1. Voting Disk Purpose:

   • Voting disks are files that help Oracle RAC determine which nodes in the cluster are healthy and should remain operational.
   • They act like a "tie-breaker" during communication failures (e.g., network issues or node failure) to prevent split-brain scenarios.

2. Split-Brain Problem:

   • Split-brain happens when nodes in the cluster lose communication with each other and both believe they are healthy.
   • If both nodes simultaneously access shared storage, it can cause data corruption.
   • Voting disks resolve this by deciding which node (or nodes) should stay in the cluster.

Why 3 Voting Disks?

Case with 1 Voting Disk:

• If you use only 1 voting disk, what happens if the disk becomes inaccessible (e.g., disk failure)?
  • The cluster cannot determine which nodes are healthy.
  • The entire cluster may shut down to avoid corruption.

Case with 2 Voting Disks:

• With 2 voting disks, if there is a communication failure:
  • Each node might claim 1 voting disk and assume it is the only active node.
  • This causes a split-brain scenario because both nodes think they own the cluster.

Case with 3 Voting Disks:

• When there are 3 voting disks, the cluster can tolerate the failure of 1 disk or the failure of communication between nodes:
• A node needs to access more than half the votes (majority) to stay in the cluster.
• If a node cannot access the majority, it is evicted.

---

Examples

Scenario 1: Normal Operation

• Node A, Node B, and 3 voting disks (VD1, VD2, VD3) are operational.
• Each node can access all 3 voting disks, and everything works fine.

Scenario 2: Network Failure Between Nodes

• Network heartbeat fails between Node A and Node B.

• Both nodes try to access the voting disks to decide which one stays.

  • Node A accesses VD1 and VD2 (2 votes).
  • Node B accesses only VD3 (1 vote).
  • Outcome: Node A wins (majority), Node B is evicted.

Scenario 3: Voting Disk Failure

• Suppose VD3 fails, leaving VD1 and VD2.
• Nodes can still function because:
  • Node A and Node B can both access VD1 and VD2 (majority).
  • The cluster continues running without interruption.

Scenario 4: Node Failure

• If Node B fails completely:
• Node A can still access VD1, VD2, and VD3 (majority).
• The cluster continues with Node A.

---

Why a Majority Is Critical

Oracle RAC requires more than half of the votes (majority) to prevent split-brain:

• In a 3-vote setup, 2 votes are needed for a majority.
• This ensures:
• If 1 node or disk fails, the other node/disk can still form a majority.
• If more than half the votes are unavailable, the cluster shuts down to avoid corruption.

---

Key Points About Voting Disk Design

1. Odd Number of Voting Disks:
   • Always configure an odd number of voting disks (3, 5, 7, etc.) to prevent tie scenarios.
2. Redundancy:
• Voting disks are typically stored on shared storage (e.g., ASM) with redundancy to handle hardware failures.