Veeam Backup Repositories and iSCSI Jumbo frames

[unsichtbarre]

I use iSCSI to mount LUNs to repository servers in two ways:

1. Using the Microsoft iSCSI Initiator to mount the disk and format is with ReFS 64K on a VM using the vmxnet3 network adapter and 10GbE LAN
2. Using open-iscsi on Ubuntu 20.04LTS to mount the disk and format it with XFS 4K on a VM using the vmxnet3 network adapter and 10GbE LAN

I always test connectivity for contiguous frame transmission from repository with: ping -l 8972 -f <IP of SAN>

Are there any disadvantages of Jumbo Frames between Veeam Repository server and the backing LUN?

THX,
-JB

[HannesK]

Hello,
if everything configured correctly, there are only advantages. So the answer to your question is "no" from my point of view.

Best regards,
Hannes

[mweissen13]

I had problems with Jumbo Frames in conjunction with Windows 10. Somehow the setting gets wiped on Windows 10's feature updates. So if you are using Windows 10 as low-budet repository host, then I would advise against it.
Of course, you should use a proper Windows Server instead

[unsichtbarre]

One thing I have found with vmxnet3:
In the 'old' days, jumbo frames were on or off.
Now you need to specify a jumbo size, and the correct specification (in spite of the vSwitch being set to 9000 per VMware) is 9014. This is the vmxnet setting you need in order to be able to : ping -l 8972 -f <IP of SAN>

[colohost]

I've not had any jumbo-specific issues with Ubuntu20LTS-based proxies mounting LUN shapshots off HPE arrays for direct backup, and using them does increase throughput and decrease interrupts/sec. Am currently working on an issue with support over one unreadable (by Veeam) LUN but it's not a jumbo issue.