Today’s Topic: Cluster Shared SAN Configuration Drift
The most common way to share data between cluster nodes is through the use of multi-homed SAN storage. Inconsistent access to the SAN volumes by cluster nodes is a state in which one or more shared volumes are not mapped to one or more nodes.
Sharing is intended to guarantee immediate data availability in case of a failover, but inconsistent mapping might put failover in jeopardy.
Why Does It Happen?
The initial configuration of a cluster is typically correct. However, routine configuration changes such as adding a new storage volume or extending the cluster to additional nodes could gradually result in a configuration drift that leaves one or more shared volumes un-mapped to some of the nodes.
What Is the Impact?
In the event of a cluster failover to the passive node, data stored on an up-mapped volume will not be available, leading to downtime of any application which requires access to a database or files stored on these volumes.
How Can It Be Avoided?
There are multiple ways to minimize the risk of such configuration drift:
1. Documentation: Put in place clear and well-documented procedures for any changes introduced to the cluster configuration.
2. Training: Conduct periodic training for all involved personnel to review possible availability risks introduced by production environment modifications.
3. Automation: Implement automated auditing of your high availability environment to ensure passive node configuration is always consistent with active node configuration.
Learn more about Automated Daily High Availability Testing