Figure 2 Failover Scenario with a Single Arbitrator
The scenarios in Table 4, based on Figure 2, illustrate possible results if one or more nodes fail in a configuration with a single arbitrator.
Table 4 Node Failure Scenarios with One Arbitrator
Failure | Quorum | Result |
arbitrator 1 | 4 of 5 (80%) | no change |
|
|
|
node 1 | 4 of 5 (80%) | pkg A switches |
|
|
|
node 1, then node 2 | 3 of 4 (75%) | pkg A and B switch |
|
|
|
node 1, 2, then arbitrator 1 | 2 of 3 (67%) | pkg A and B switch |
|
|
|
nodes 1, 2, arbitrator 1, then node 3 | 1 of 2 (50%) | cluster halts* |
|
|
|
arbitrator 1, then node 1 | 3 of 4 (75%) | pkg A switches |
|
|
|
data center A (nodes 1 and 2) | 3 of 5 (60%) | pkg A and B switch to data |
|
| center B |
|
|
|
data center A, then arbitrator 1 | 2 of 3 (67%) | pkg A and B switch, then no |
|
| change |
|
|
|
data center A and arbitrator 1 | 2 of 5 (40%) | cluster halts* |
|
|
|
data center A, then arbitrator 1, then node 3 | 1 of 2 (50%) | cluster halts* |
|
|
|
arbitrator 1, then data center A | 2 of 4 (50%) | cluster halts* |
|
|
|
node 3, then data center A | 2 of 4 (50%) | cluster halts* |
|
|
|
data center B | 3 of 5 (60%) | pkg C and D switch to data |
|
| center A |
|
|
|
third location | 4 of 5 (80%) | no change |
|
|
|
* Cluster can be manually started with the remaining node.
With a single arbitrator node, the cluster is at risk each time a node fails or comes down for planned maintenance.
Example Failover Scenarios with Two Arbitrators
Having two arbitrator nodes adds extra protection during node failures and allows you to do planned maintenance on arbitrator nodes without losing the cluster should a disaster occur.
30 Designing a Metrocluster