Anyone having horrible performance in DPM 2025?

Question

Anyone else experiencing horrible performance with DPM 2025? After upgrading from the latest update rollup of DPM 2022, we have been experiencing major issues that after about a week of operation make the system unusable and requiring reboot. After reboot, the system can catch up, but then in about a week the problem is back. This is causing many recovery points to be missed. CPU use is normally pegged near 100% at this time. Increasing the number of cores beyond the recommended number of 4 does not help. Many jobs are in the queue for things such as small SQL databases that in the past would complete in less than a minute but are now running for several days. I'm running on same hardware as with DPM 2022. Backup load is essentially the same as well. Only other difference is SQL 2022 is being used instead of SQL 2019 as well as running on Windows Server 2025 instead of 2022.

Answer 1

Threw out everything and started back with Windows Server 2022. After a year since its initial release, Windows Server 2025 is simply not ready for this use of ReFS. While I had to discard everything, I was missing so many recovery points with the high CPU use that it was worth it.

Now CPU use is more normal. DPM 2025 still does not seem to perform as fast as DPM 2022, and I am still getting these frequent errors ID 3114 on SQL databases and SharePoint farm databases stating I do not have the correct OLE DB driver installed when I absolutely do (19.4.1).

Answer 2

Unfortunately, after applying the August update (KB5063878), as hard as it is to believe, I actually think behavior is worse. After this update, most of my SQL data sources are failing with errors like:

Affected area:	SQLServerName\DBName
Occurred since:	8/20/2025 12:49:23 AM
Description:	Recovery point creation jobs for SQL Server 2022 database SQLServerName\DBName on sqlservername.domain.com have been failing. The number of failed recovery point creation jobs = 31.
 If the data source protected has some dependent data sources (like a SharePoint Farm), then click on the Error Details to view the list of dependent data sources for which recovery point creation failed. (ID 3114)
	An unexpected error occurred while the job was running. If backup is failing for a SQL instance or Sharepoint Farm, make sure SQL OLEDB driver version 19.3.3 or higher is installed on protected computer. (ID 104 Details: The parameter is incorrect (0x80070057))
	More information
Recommended action:	Retry the operation. If backup is failing for a SQL instance or Sharepoint Farm, make sure SQL OLEDB driver version 19.3.3 or higher is installed on protected computer.
	Create a recovery point...
Resolution:	To dismiss the alert, click below
	Inactivate

This happens within a few hours of reboot. Everything seems to be stable for a few hours, but then most of the databases and my SharePoint databases give this error. And yes, I am running OLEDB driver 19.4.1 as I was before the update was applied. The SQL servers are a combination of 2019 and 2022, but they are all running on Windows Server 2025. I'm not certain if the problem is on the SQL end or the DPM end as the SQL servers are issuing many WMI event 5858. I also installed the updates on the SQL servers.

At least in the past, I could reboot the system every few days and return to essentially normal behavior. Now, I cannot get past these errors, and after a few days of running the performance is horrible again (the errors persist during this entire time).

Answer 3

I found this article last week and it describes my issue perfectly:

https://windowsforum.com/threads/windows-server-2025-refs-bug-critical-impact-and-deployment-strategies.372652/

I've had a support case open on this for almost 6 months with no progress, just a lot of DPM log gathering and phone calls. I shared this link with Microsoft, but they are hesitant to state that this is the problem. Since I did not seem to have this issue on my secondary server, I decided last week to reinstall everything from scratch using the May 2025 image of Windows Server 2025 followed by bringing it up to current July updates. I retained my DPM database and storage pool.

It's only been a week so my opinion may change, but so far everything seems to be closer to normal with more reasonable CPU use and expected disk or network activity. I'm not seeing CPU use constantly in the 90% range on a 4-core system (what Microsoft recommends for hardware) with almost no disk or network activity. Before the reinstall, I would see CPU use in that range within a couple of days with rebooting the only solution to getting it back to normal. I didn't see memory use pegging, but I have much more than the minimum recommended amount in this system. I truly believe the issue described in the article is the root problem, especially since it was found with backup software as well. We've had tons of problems with Windows Server 2025 (24H2) as well as Windows 11 (24H2), more than any other version going back to Windows NT4, so it is not hard for me to believe this could be another bug.

A couple of months before upgrading to 2025, I did have a problem with 2022 where the storage was being unexpectedly bloated because the pruning jobs were not properly working.