Questions around S2D on GCP RRS feed

  • Question

  • We are having a performance issue with GCP. We have setup a FCI cluster on S2D following the documentation on GCP around creating an HA MS SQL cluster. The cluster is validated, and everything works as expected: SSMS access to SQL Server databases on the FCI, cluster managment, failover operations, etc., but we are witnessing very odd behavior when doing load testing. During the load test we are running against the DB cluster, one server in the cluster only hits 40% cpu utilization, but when we failover to use the other server in the cluster, that server goes straight to 80% CPU usage. The load being applied to the cluster remains constant over this time. Even more odd, we can make this behavior occur on either server. Whichever server was the last to do a full stop start from the GCP console will be the server that has low CPU when active. Our current thought is this might be a network latency issue with S2D, but we could use assistance in understanding the cause of this behavior. We would like help identifying and resolving the performance issue.
    Wednesday, September 25, 2019 6:25 PM