Speakers: Josh Gray, Jeff Hunter
This was a great session for real-world HA tips and tricks of the trade. Everyone should be using vSphere HA, so theh tips apply to everyone. For more information check out the most excellent vSphere 5.1 Clustering Deepdive book by Duncan and Frank.
- Recent Enhancements in vSphere HA
- Fault Domain Manager released in v5.0. Completely re-written from scratch.
- No more reliance on DNS; just uses IP communications
- Datastore heartbeat added
- vSphere 5.1: FDM agent included in image, no more manual autodeploy configuration. Can manually set slot size. Enhanced PDL (permanent device loss) handling.
- HA Networking Recommendations
- Redundant management network
- Fewest hops possible
- Route based on originating port ID (don’t enable Etherchannel)
- Failback policy = no, if PortFast is not enabled
- Enable PortFast, Edge
- MTU size the same
- Keep things simple
- Consistent porgroup names, network labels are consistent
- Host monitoring during network maintenance
- Use maintenance mode
- Separate subnet for vSphere HA
- Specify additional network isolation address, specially if default gateway will not return pings
- Advanced settings: das.AllowNetwork, das.isolationAddress, das.usedefaultisolationaddress
- HA Storage Recommendations
- HA selects two datastores by default
- Override auto-selected datastores only when needed (e.g. there is one highly available array and one that’s less available)
- Host Isolation Response
- Policy depends on what is most likely to fail, or retain access to (see slide below)
- Admission Control
- Static Number of Hosts – Sets aside the amount of resources for a single host.
- vSphere 5.1 allows you to explicitly set the slot size, and re-words the setting for host cluster failures tolerates setting.
