Backup to CIFS Repo on NetApp filer painfully slow

[mdiver]

With one customer we observe backing up to a NetApp filer is painfully slow (~100kB/s).
The filer is mapped as CIFS repository to a datamover sitting inside a VM on one of the hosts. NBD is used as tranport mode to the proxy.

To test the repository role, a second repository is mapped to the very same VM from a QNAP filer. This one shows reasonable rates of >50MB/s. So the repo VM seems to be fine.

Also if we copy files manually to the NetApp share being used as CIFS targe from the same VM via Windows-Explorer we get rates of >50MB/s as well.
So only via Veeam and the repository role the filer is accessed with this painfully low speed.

A support case is filed as #02138631 and already on 3rd Level. But no ideas so far.

Could it be some special blocksize issue or Setting with the NetApp? Anyone else had this kind of issue?

Thanks an regards,
Michael

[foggy]

Hi Michael, I'm sure support has already checked all that, but still, what are the bottleneck stats for this job and are both storages similar in terms of network topology (located at the same site, use the same datapaths, etc.)?

[mdiver]

Hi Alexander.

The backup is running at a remote site in respect to the Veeam B&R Server.

Though both repositories tested (NetApp as well as QNAP for comparison) are CIFS targets in the same subnet on the remote site.
Also the ESX-host, the proxy used and the gateway Server are at the remote site in the same subnet.

The only difference is the CIFS target used. Using the NetApp slows down the speed by some orders of magnitude compared to the way cheaper QNAP.
As said: When we copy any large file directly using Windows Explorer from the same gateway server, the speed is fast and almost the same for both Targets (1G-Ethernet).
So only through Veeam's DataMover we get the behavior.

We just upgraded to 9.5U2 as requested by support. No changes.

Bottleneck is always 99% target. Sometimes also 99% network (which does not make any sense as proxy and gateway are on the same VM).
But mostly 0-2% for everything except "target".

Thank you,
Michael

[foggy]

Everything points to the storage itself.

[skrause]

Is the NetApp by chance close to being full?

[mdiver]

Together with Veeam support we found a solution.
Setting this regkey on the proxy/gateway brings us the same (reasonable) speed for the NetApp target as we already had to the QNAP target.

Key: HKLM\SOFTWARE\Veeam\Veeam Backup and Replication\
DWORD: NetUseShareAccess
Value: 1

This leads to the gateway using a resilient method for connecting the CIFS target.
It seems the NetApp doesn't like the other method.

Regards,
Mike