Failed to flush change tracking data / Slow backups

[jpirog]

Case #04878815

As I've submitted in the support case, I don't know if this is normal or not but our backups seem to be very slow all around. I've only posted one of the jobs for one VM but it's the same across the board.

We are running on a HyperV/Nutanix Cluster currently. The Veeam BR server is a separate physical server but connected to the same fiber 10Gbit switch.

I had posted on another forum and everyone had mentioned CBT wasn't working correctly and that would cause backups to be slow. But it also makes me wonder if there's anymore that can be done to speed everything up. The overall network bandwidth really shouldn't be an issue considering it's on a 10gig fiber backend.

[veremin]

Sometimes it happens when one cluster node gets paused. To fix the issue you can try to reset CBT, using script from this thread. Or you can wait for support engineer answer and let him or her assist you with the investigation. Thanks!

[jpirog]

Awesome, thanks veremin, I'll take a look at that and see if that will help fix the issues.

What about the disk reads? Is it normal for each individual read to only be about 15-30MB/s?


Edit, getting this error:

Code: Select all

Reset-HvVmChangeTracking : Cannot validate argument on parameter 'Server'. The argument is null. Provide a valid value
for the argument, and then try running the command again.
At line:1 char:34
+ Reset-HvVmChangeTracking -Server $Server
+                                  ~~~~~~~
    + CategoryInfo          : InvalidData: (:) [Reset-HvVmChangeTracking], ParameterBindingValidationException
    + FullyQualifiedErrorId : ParameterArgumentValidationError,Veeam.Backup.PowerShell.Cmdlets.ResetHvVmChangeTracking

[soncscy]

Your powershell error is just that $Server is a null value, so when you ran Get-VBRServer, it didn't find whatever you fed it. Try

$Server should be fed like:

$Server = Get-VBRServer -name 'the name of your server as it shows in Veeam'

You can use the -Type parameter also to specify HyperV Host. (Just add -Type then a space, then press tab to go through the options)

>What about the disk reads? Is it normal for each individual read to only be about 15-30MB/s?

Normal is relative to your environment HyperV hosts themselves are the proxies for reading disks; confirm the bottleneck in the job statistics shows as Source, and if so, you can use Diskspd on a copy of a VHDX file on your host.

https://www.veeam.com/kb2014

The Active Full/Forward Incremental should be fine, but you will want to compare it to a similar run when the host isn't doing much else.