SQL Server Availability Impact on BigBox – Non yielding Resource Monitor after changing NUMA-NODE Processor Affinity causes instance to failover

No comments

Change NUMA-Node Processor Affinity is an online operation that can be done without causing downtime but I have experienced that it can lead to bring an instance offline if the resource monitor takes long time to free memory and keeps generating Mini-Dumps, at #51 it will bring the instance offline.

Started to have Non yielding Resource Monitor after changing NUMA-Node Processor Affinity in a Microsoft SQL Server 2014 (SP1-CU5) instance.

Looking at each stack for threads which have a time > than the Time the NUMA-Node change was done there is a thread that keeps increasing User Mode Time with the following stack, that is related with freeing physical pages:

Think twice about changing NUMA-Node Processor Affinity if your instance isn’t using large page allocations TF834 , activating this flag will change code path and make use of LargePageAllocator 


Paulo Condeça.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s