Title here
Summary here
How High Availability Works
Overview
High Availability (HA) detects a compute resource failure through the shared-storage lock, not through the SolusVM agent. This section describes the recovery process step by step and the conditions that must apply before HA starts an evacuation.
The Recovery Process
The following steps describe how HA protects a virtual server when a compute resource fails:
- Compute resources in the failover domain continuously report their shared-storage lock status.
- SolusVM confirms a compute resource failure when the other compute resources in the failover domain confirm that its storage lock has expired.
- After the grace period expires, SolusVM places the affected virtual servers on the healthy compute resources in the same failover domain. It places each virtual server individually. They may all land on one compute resource or spread across several, depending on available resources.
- SolusVM places each protected virtual server on the selected compute resource. The virtual server disks stay on the shared storage and do not move.
- If a virtual server was running when the failure occurred, HA starts it on the new compute resource.
- The virtual server keeps the same IP address, so services remain available through the same network address after a failover.
- When the original compute resource comes back online, the virtual server continues to run on the new compute resource. You must manually move the virtual server back to the original compute resource.
Note:
- HA does not rely on the SolusVM agent or the connector between the management node and a compute resource to confirm a failure. If the management node loses network access to a compute resource but that compute resource still holds its storage lock, HA considers it active and does not start an evacuation. This prevents false failover events when a network connection fails, but the compute resource and its virtual servers continue to run.
- HA and Disaster Recovery can coexist. If HA cannot migrate a virtual server and that virtual server has a valid backup, you can perform Disaster Recovery manually.
Evacuation Triggers
HA starts an evacuation event automatically when all of the following conditions apply:
- The other compute resources in the failover domain confirm that the storage lock of the affected compute resource has expired, which means its watchdog has fired.
- The confirmed failure persists for longer than the grace period.
- The compute resource completed its failover domain join successfully, and it is not in maintenance mode.
- The compute resource belongs to an active failover domain.
- The HA global setting is enabled.
To review all HA evacuation events, see Monitoring High Availability Activities.