Case 01108056 - CBT Crash

[PTyrer13]

I have been advised by Stanislav Artemov a member of Veeam B&R support team to create a post on the forum regarding the issues that have arisen from the above case. In short a power cut occurred and our VM hosts went down, when they were back up and running and the backups started to process they finished and reported as successful. On an attempt to restore a file on the following day it was noticed that the backups from the previous evening had not picked up files which we knew to have been changed. It was found by the above support agent that the CBT had crashed on our hosts and need to be manually restarted. What concerns us is that no error or warning from Veeam appeared to tell us the CBT had failed and the read and transferred data where not the same value which the above agent had mentioned was a warning sign of such a problem.

Any insight would be much appreciated.

[foggy]

Paul, are you positive that the file was changed prior the power cut?

Basically, in case of the host crash, CBT is reset automatically and during the next job run entire disks should be read. Read counter should not be equal to Transferred, since still only changes are transferred in case, just the Read counter is high due to the need to re-read entire VM data and calculate differences (it's not a full backup). Also, I'm not sure about what should be manually restarted, since everything is automatic in this case.

[Cragdoo]

Apologies for adding to the thread. Does the CBT reset also occur when a host is placed into maintenance mode?

We recently put a host in our Hyper-V cluster into maintenance, then powered down the host. On the next backup run, the backups took considerably longer. This was due to missing CBT information from the host we powered down.
As I understand it, due to the lack of a native CBT mechanism in HyperV , Veeam created their own, which is 'injected into each host of a cluster.
Am I right is saying this information is not replicated across the cluster hosts? So in the event of a host failure, VMs are migrated to another host, but the CBT information remains on the original host ??

[foggy]

Craig, you're correct. CBT is also reset in case of placing the host into maintenance mode. Your understanding of how CBT works is also correct.

[Cragdoo]

Thanks for confirming foggy.

Are there any plans to introduce distributed CBT tracking, so placing a host in maintenance or host failure wont reset the CBT? This is kind of the case at the moment with VMware backups, as the CBT data is placed with the VM (which usually resides on shared storage)

[foggy]

At least not short-term plans. There's a slight hope for native mechanism in Windows Server 2016, though.

[PTyrer13]

Hi Foggy, Yes is the answer to the first question, I myself created a new document on one of our VM's around 11.30am and the cut did not occur until 4.30pm. The second part I may not have worded correctly, Basically I had to follow the steps in the bottom half of support article KB ID: 1934 "Changed block tracking can be reset for a Hyper-V cluster by performing the following steps, should the information above not resolve the issue. To reset changed block tracking, please perform the following: "
Paul

[foggy]

I recommend contacting technical support for a more close look at your environment. According to your description, the new document should be available in the restore point created after the power cut.