Failover Cluster

Window-Server-Feature-Image

Failover Cluster
1. A group of comupters work together.
2. Each independent computer calls node.
2. It can provide High availability and scalability.
3. It will have its own hostname and IP
4. It is recommanded to use multiple NIC
5. It is recommanded to separate Client network and Cluster network.

Split Brain
1. It is a situation which the nodes lose connections between each other.
2. It will causes many porblem such as unresponsed node, Storage corruption.

Quorum
1. It is used to prevent split-brain
2. It is a voting algorithm
3. Each node in the cluster has a vote
4. The cluster still works when more than half of the nodes are online
5. If there are an even number of nodes and no witness, one nodes gets its vote zeroed to keep the total number of voes is odd.

Witness
1. It is used to confirm the total number of votes is odd
2. It also has a vote like node if there are an even number of nodes
3. If there are an odd number of nodes, the witness will not vote

Some example of Quorum:
Example A:
1. Total number of nodes is 2
2. No Witness
3. One of nodes gets its vote zeroed, the total number of voter is 1
When 1 of the nodes is down, then the cluster will down.

Example B:
1. Total number of nodes is 2
2. With witness
3. The total number of voter is 3
When 1 of the nodes is down, then the cluster will still online.

Example C:
1. Total number of nodes is 3
2. No Witness
3. Since total number of nodes is even, no nodes get its vote zeroed
When 1 of the nodes is down, then the cluster will still online.
Beside, the situtation of cluster beomce EXmaple A.

Example D:
1. Total number of nodes is 3
2. With witness
3. Since total number of nodes is even, witness get its vote zeroed
When 1 of the nodes is down, then the cluster will still online.
Beside, the situtation of cluster beomce EXmaple B.

Quorum witness types
1. Cloud Witness uses Blob storage in Azure as witness
2. File Share Witness uses SMB file share folder as witness

Lab information
Failover Cluster Network Diagram

Network preparation
1. Two NICs per Server:
One for cluster network
One for iSCSI network

AD DC preparation
1. Create new OU (CLUSTER) for the cluster computer objects

2. Create computer account (CST00) in new OU (CLUSTER) which assign as cluster name
3. Disable the computer account

4. Grant the computer object (CST00) “Create Computer objects” on the OU (CLUSTER)

5. Create domain user account (st-admin) who installs the cluster
6. Add the user to “Account Operators”
7. Grant the user (st-admin) “Full Control” on the computer account (CST00)

Node server preparation
1. Add domain user account (st-admin) in local administrator group
2. Install “Failover Clustering” feature

Setup Failover Cluster
1. Verify Cluster hardware and settings

2. Create Cluster

3. Verify Cluster hardware and settings

Create File Share Witness
1. Create share folder on ST00
Folder Path: C:\Witness
Shared Folder Name: “CST00-Witness”
Grant “modify” Permission to “joeschoice\CST00”

2. Create Witness on ST01/ST02

Remove File Share Witness

View Cluster Heartbeats
1. CrossSiteDelay — Heartbeats interval, in milliseconds, across sites
2. CrossSiteThreshold — How many heartbeats can be missed across sites
3. Default value of CrossSiteDelay is 1000 and CrossSiteThreshold is 20

4. CrossSubnetDelay — Heartbeats interval, in milliseconds, across subnets
5. CrossSubnetThreshold — How many heartbeats can be missed across subnets
6. Default value of CrossSubnetDelay is 1000 and CrossSubnetThreshold is 20

7. SameSubnetDelay — Heartbeats interval, in milliseconds
8. SameSubnetThreshold — How many heartbeats can be missed
9. Default value of SameSubnetDelay is 1000 and SameSubnetThreshold is 10

Change Cluster Heartbeats

Remove Cluster
1. Run the following PowerShell in ST01

Author: Joe Chan

2 thoughts on “Failover Cluster

Comments are closed.