| Question |
Answer |
Further reading |
| When does failover occur? |
- Failover occurs when a quorum of nodes votes a node out of a cluster
- Voting depends on timely cluster heartbeats between its nodes. If a node isn't communicating with other nodes in the cluster, it gets voted out and its forests are failed over
|
KB Articles:
Documentation:
|
| What nodes participate in quorum? |
All nodes in the cluster configuration count towards quorum, irrespective of the:
- group they belong to
- type of node it might be (E-node, D-node, E/D-node, etc.)
- state of the node (online/offline)
- forest, database or group configurations
|
KB Articles:
Documentation:
|
| My cluster saw a failover event - does fail back happen automatically? |
- Failing back is a manual operation
- Automatic fail-back is not supported due to the risks of unintentional overwrites and accidental data loss
|
KB Articles:
Documentation:
|
| How should I distribute my forests across the nodes in my cluster? |
- Here is an example forest topology on a typical 3-node cluster:
- Node1: df1.1, df1.2, ldf2.1, ldf3.1
- Node2: df2.1, df2.2, ldf1.1, ldf3.2
- Node3: df3.1, df3.2, ldf1.2, ldf2.2
- Distributing local disk failover forests (LDFs) evenly on the other two nodes splits the load on each surviving node to just 150% of normal should a failover event occur
- When scaling out the cluster, try adding nodes in "rings of three" to keep the above load distribution intact with minimal config changes within each ring (for example, nodes 4, 5, and 6 would mirror the data and LDF forest distribution seen in the ring made up of nodes 1, 2, and 3)
|
KB Articles:
|
| How many Local Disk Failover forests (LDFs) should each of my primary forests have? |
- One LDF for each primary forest, hosted on a different machine than the one which hosts the primary forest
- More than one LDF per primary is not recommended due to unnecessary increases in administrative complexity and hardware resource requirements
|
KB Articles:
|
Should I be backing up my local disk failover forests?
|
- Local disk failover forests (LDFs) should be included in your backup if you expect backups to be taken in a failed over state. Note that this will typically double the size of your backup since you're including both data and LDF forests
- If you're not in a failed over state, or you manually fail back before a backup starts, then you can reduce the size of your backups by only backing up your data forests
|
KB Articles:
Documentation:
|
(16 vote(s))