Problem Statement

To enhance the isolation detection abilities of vSphere to minimize the chance of false positive isolation responses  Datastore Heartbeats will be used. What is the most suitable configuration of Datastore Heartbeats for an environment using SRM?

Requirements

1. SRM solution must not be impacted

2. Maximum vSphere environment availability

Assumptions

1. Site Recovery Manager 5.1 protects virtual machines in the cluster/s

2. Appropriate isolation address/es have been configured OR the default isolation address is suitable

3. As all storage is presented via Active/Active storage controllers

4. There are some datastore which are not replicated

5. Isolation response is set to “Shutdown”

Constraints

1. None

Motivation

1. Minimize the chance of a false positive isolation event

2. In the event of isolation, automate the recovery of VMs

Architectural Decision

Use Datastore Heartbeats to enhance the isolation detection capabilities of vSphere.

For each cluster where SRM is used, Configure Datastore Heartbeating to Manually select two non replicated datastores per cluster as the heartbeat datastores

Justification

1. Datastore heartbeating frequently writes to the datastore selected for heartbeating so in the event the network is down, isolation, partition or failure can be properly determained. As a result, during a SRM recovery, datastores need to be un-mounted from the failed site and the Datastore heartbeating may cause one or more datastores to fail to unmount due to I/O on the datastore

2. Datastores failing to un-mount will cause one or more of the SRM recovery steps to report as failed, selecting non replicated datastores prevents this impacting SRM

3. The environment benefits from increases resiliency as a result of datastore heartbeats being used

4. There is no negative impact to the SRM solution

Implications

1. Each cluster will need to have one or more non replicated datastores if Datastore Heartbeating is to be used

2. Additional configuration required to manually select non replicated datastores for heartbeating

Alternatives

1. Do not use Datastore heartbeating

2. Use Datastore Heartbeats and have datastores automatically selected

Relates Articles

1. Example Architectural Decision – Host Isolation Response for FC Based storage