In fourth post of HP StoreVirtual VSA series we are gonna discuss about storage clusters composed by two or more virtual storage appliances.
As said in previous articles HP StoreVirtual is a scale-out storage solution which benefits from the addition of other nodes by increasing data reliability and throughput by load balancing requests through different nodes.
To create a proper cluster with load balancing features we need at least two VSAs.
When dealing with a cluster comprising two or more HP StoreVirtual VSAs we need to introduce a new component named as Failover Manager (simply referred as FOM). As for the physical HP StoreVirtual counterpart even VSA use a Failover Manager. Basically FOM is a virtual appliance deployed in VMware environment and to quote official HP documentation:
A specialized manager running as a VMware guest operating system that can act as a quorum tie-breaker system when installed into a third location in the network to provide for automated failover/failback of the Multi-Site SAN clusters. It also added to two-system management groups to provide the optimum quorum configuration.
The Failover Manager is a specialized version of the SAN/iQ software designed to run as a virtual appliance in either a VMware or Microsoft Hyper-V Server environment. The Failover Manager participates in the management group as a real manager in the system; however, it performs quorum operations only, not data movement operations. It is especially useful in a Multi-Site SAN configuration to manage quorum for the multi-site configuration without requiring additional storage systems to act as managers in the sites.
Despite HP StoreVirtual could even use the embedded Virtual Manager to manage quorum ownership in a single-site configuration it is a common practice to deploy FOM due to the fact that the former has some known limitations.
FOM is deployed as a virtual appliance either installable from the HP StoreVirtual VSA for VMware Installer, either downloadable from HP website as an ova assembly. In this scenario I used the former solution.
Since FOM need to communicate with all VSAs in the cluster a proper IP address must be assigned to it.
Once FOM has been installed it has to be found using CMC as for a common VSA by inserting its IP address.
Regardless you are going to create a new cluster or adding FOM to an existing one, the optimal configuration is based on one FOM per cluster plus two, or more, VSAs.
A major benefit brought by clusters with two VSAs is the ability to use Network RAID-10 in which data is replicated on both VSAs to provide a greater degree of availability.
In case of outage of the physical ESXi host running one VSA data is still available and online from other VSA running on the other powered-on host.
When VSA on failed host will be powered back on by HA it will be automatically recognized online and will actively partecipate in the cluster again.
When three or more VSAs are added to a cluster the Data Protection Level of volumes becomes pretty intresting. Network RAID-5 and Network RAID-10+1 can also be selected.
Network RAID-5 is the least space consuming protection level which only distributes parity across VSAs while Network RAID-10+1 is the most secure, yet the most space consuming, level available of data protection which replicates volume data on all three VSAs vaulting data against two appliance failures at the same time.
In a two VSA cluster the cluster itself is not impacted in case of FOM failure.
In case of failure of both FOM plus one of the two VSA the cluster becomes unusable until FOM and/or VSA are brought back on.
To prevent this situation a good practice is to place every appliace (FOM + both VSAs) on separate physical hosts decreasing the probability that an host outage could affect the entire storage cluster.
Other blog posts in HP StoreVirtual VSA Series:
HP StoreVirtual VSA Part1 - Installation
HP StoreVirtual VSA Part2 - Initial Configuration
HP StoreVirtual VSA Part3 - Management Groups Clusters and Volume
HP StoreVirtual VSA Part4 - Multi VSA Cluster