Tirning off the usage of Buffer Pool Extension

Question

We are using the SQL Server 2016 Standard with Buffer Pool Extension enabled on SSD drive.
For external nightly maintenance we need to turn off an extensive writing on this SSD.

So we have scheduled the turning off BPE. Checked it on the testing instance.
But our production SQL Server instance began to hang with memory allocating errors in logs just after the ALTER SERVER CONFIGURATION script run.
BOL doesn’t say anything about mandatory instance restarting after turning off the BPE. And we also don’t want it to be restarted.

It’s unclear why it can’t handle with changing of BPE without reset, while it contains only the “clean pages”? Even size of available RAM can be recofigured online without the issues.

So, should be BOL fixed about this issue?
Does there any other way to temporarily freeze the disk writing to BPE file (we have seen up to 25 Gb per hour in non-production time) ? (I have tried to use 861 and 862 trace flags, but according to the performance counters BPE writing doesn't stop)

Answer 1

Sounds to me like a bug. I know it might be difficult but either update to current sp/CU (this might have been fixed already) or see if you can repro it and report to MS. You can of course open a case with MS even without a repro, but a repro will most likely shorten the time for a solution.

Answer 2

Hi 38416426,

In addition, you can submit this issue to the Microsoft feedback at this link: https://feedback.azure.com/forums/908035-sql-server.
Your feedback enables Microsoft to offer the best software and deliver superior services, meanwhile you can learn more about and contribute to the exciting projects on Microsoft feedback.

Best Regards,
Amelia