FlashArray ActiveCluster for Microsoft SQL Server Failure Scenarios

This section examines the availability of storage and network resources in different failure scenarios, and it also discusses troubleshooting common issues.

ActiveCluster Component Failures

Table 1 lists storage failure scenarios that can occur with ActiveCluster, and how data availability to hosts is affected.

Table 1. ActiveCluster component availability scenarios
Components				Storage Availability	SQL Server
Array 1	Array 2	Replication Link	Mediator	Storage Availability	SQL Server
Up	Up	Up	Up	Available on both arrays	Up and running on all the hosts, with access to both arrays
Up	Down	Up	Up	Available on the surviving array	Up and running on all the hosts, with access to Array 1 only
Up	Up	Down	Up	Available on one array	Up and running on all the hosts, with access to the chosen arrays
Up	Up	Up	Down	Available on both arrays	Up and running on all the hosts, with access to both arrays
Up	Down	Down	Up	Available on one array	Up and running on all the hosts, with access to Array 1 only
Up	Up	Down	Down*	Unavailable	Down, as both the replication link and the mediator are down
Up	Down	Down	Down*	Unavailable	Down, as both the replication link and the mediator are down, along with an array
Down	Down	Not applicable	Up	Unavailable	Down, as both of the arrays are down

^*Entries in Table 1 labeled with asterisks (*) denote failures of components other than inaccessibility of the mediator. If the mediator becomes inaccessible after an array or the replication link has already experienced a failure, storage remains accessible on one array.

Host and Storage Network Failures

Table 2 lists the observable behavior of select storage components in the event of various failure scenarios.

Table 2. Storage component failure scenarios
Failure Scenario	Behavior
Single or multiple host failure	Applications can automatically failover to other hosts on the same site or to other hosts at another site connected to the other array. This is driven by the Windows Server Failover Cluster feature assuming clusters are stretched between sites.
Stretched storage-area network fabric outage (Fibre Channel or iSCSI) (Failure of storage-area network interconnect between sites)	Host input/output continues automatically on local paths within the local site: Encounters some storage path failures to the remote array but maintains input/output on paths to the local array. In each site, retains access to local volumes with only a brief pause in input/output.
Storage-area network fabric outage at one site	Applications can automatically switch to hosts at another site connected to a different array, driven by the host SQL Server cluster: In the site without a storage-area network issue, there might be some storage path problems for paths to the remote array, but operations traffic continues through paths to the local array, The volumes on the remote array would be accessible, but latency would increase. The administrator could failover the cluster to a node in the unaffected site.