I am trying out Veeam B&R (version 6 with patch 2) very briefly so far, I've only added one backup job containing one Windows 2008 R2 VM. The VM has two 50GB VMDKs and the backup job is set to include both of them, backing up to a CIFS repository, reverse incremental, dedupe, no compression, no VMware Tools quiescence, *use CBT* and do application-aware image processing. The ESXi host storage is local DAS.
I added it one or two days ago, and for the first eight runs the backup completed just fine. I then tried an instant VM recovery, by having the backup controller publish it the VM from the backup via its own NFS (vPower NFS I think?). This worked fine too, the backed up VM was now running on the ESXi host. At first I ran it without having it connected to the network, and along side the "real" one, but then I shut down the real one and connected this backup one to the network to test it a little. So far so good. The backup job complained that it could not back up the VM anymore due to some files being locked, which I think is fine.
So after having tried this I then right-clicked the Instant Recovery VM and selected "Stop publishing". Veeam shut down the backup VM on the ESXi host, all fine. However on the following backup runs of the "real" VM, I got warnings: One saying
Code: Select all
2012-01-27 05:34:41 :: Hot add is not supported for this disk, failing over to network mode...
Code: Select all
2012-01-27 05:34:45 :: Cannot use CBT: Soap fault. A specified parameter was not correct. . deviceKeyDetail: '<InvalidArgumentFault xmlns="urn:internalvim25" xsi:type="InvalidArgument"><invalidProperty>deviceKey</invalidProperty></InvalidArgumentFault>', endpoint: ''
I looked in the forum and found some reports on the CBT issue, for example this. While this does seem relevant, it describes an issue where the CBT problem happens only on the second and later VMDKs, not the first one which I am seeing.
In any case I did shut down and restart the VM a couple of times, and after that I no longer have both of the above warnings, but only the Hot Add warning. To clarify; the current backup runs that are made for the VM only give the "Hot add is not supported for this disk" warning, but still for both VMDKs. I have tried to make a snapshot of the VM, and then deleted all of them (though there's only one) as I saw mentioned in another thread, but the warnings persist (and indeed the Hot Add does not work, it backs up over network now).
There is one more thing to note however, namely that after I stopped the publishing of the backed up VM and Veeam shut it down, I saw some heavy disk activity for it on the ESXi host, for quite some time. I do not know what that was about, as I had not asked Veeam to finalize/migrate the backed up VM to the ESXi host. However, the write cache during the Instant Recovery run was set to the ESXi host local storage, so could have something to do with that. I do presume, however, that the changes I made to the backed up VM during its test run were not saved back to the actual backup of it.
Any ideas why this might be and how to fix it without a new full backup (something I can do now but would like to not do in the future when there's a lot more data)? The general observation is that it's started since I did a simple Instant Recovery test run.
I know that Veeam are working on a workaround for a CBT problem that apparently is in ESXi/VMware/vSphere, but I'm not sure if it 1) will also fix this problem I was seeing for *both* VMDKs, and 2) has to do with the Hot Add problems as well.