Home > Storage > PowerFlex > White Papers > Dell APEX Block Storage for AWS: Microsoft SQL Server 2022 on Linux with Multi-AZ Resiliency > Validate AZ failure
This section provides a detailed description of the test performed to verify the storage resiliency of SQL Server 2022 with Dell APEX Block Storage for AWS.
The test was performed to evaluate storage resiliency while the SQL Server 2022 cluster has active transactions using the HammerDB tool. HammerDB is sensitive to disconnections from the workload, any such events result in an error and stop the test.
The following Figure 17 shows the APEX Block Storage and SQL server setup used for testing storage resiliency in AWS public cloud.
To simulate a real-life scenario, SDS nodes running in AZ-1C were forcefully shut down while the HammerDB tool continued to generate transactions.
HammerDB was used to generate database transactional workload on the cluster and when the workload was running in a steady state, the SDS nodes which were part of AZ-1C were powered off from the Azure Portal. This allowed us to simulate an AZ failure, as shown in the following figure:
When the Dell APEX Block Storage cluster detects the failure of nodes on AZ-1C, the self-heal capability initiates the rebuilding process of the volumes’ data that was affected by the failure, as shown in Figure 19. PowerFlex Manager shows a temporary dip in IOPs while the rebuild is in progress, and then the system regains its original performance, while the HammerDB and SQL Server continue operations. The following figure shows the PowerFlex Manager dashboard:
Storage IOPS and HammerDB transaction rates return to normal after the APEX Block Storage rebuilding process is completed. During the failure event and the rebuilding process, the application (HammerDB) had no impact and continued to work as expected.