Short synopsis: "There is an issue with Change Block Tracking (CBT) in Express Patch 6 for ESXi 6 (Build 3825889) that can affect your backup if your backup software relies on CBT."
Affected are Windows VMs on HW version 7 to 11 and with any VMware Tools version on both thin and thick provisioned disks while doing backups with quiesced snapshots.
Solutions so far:
don't use quiesced, i.e. application-aware snapshots
or downgrade to a older ESXi6 version
If you have been hit, you have to reset your CBT data after fixing the problem (as usual with CBT problems).
Does somebody know already if this means corrupted backups or only slow backups because CBT is just not working? From the article it sounds like CBT is just not working and the backups should be ok...
mkretzer wrote:Does somebody know already if this means corrupted backups or only slow backups because CBT is just not working? From the article it sounds like CBT is just not working and the backups should be ok...
Can anybody please confirm that (other than the last CBT bug) this newly introduced bug doesn't mean corrupt backups which B&R has no way to detect? While extended backup time due to unavailability of CBT is bad, unnoticed backup corruptions are of course much worse. Thanks.
I'm still running patch 3380124 (Update 1b) on my newer hosts, as I think this was the only patch that had no real issues.
I've updated vcenter to U2, but this whole upgrade project seems to be dragging on forever as there's no new build that is good and I'm too wary of upgrading my good 5.1 hosts.
With the mess that was 5.5 and its patches and updates that caused purple screens, most of my customers are still running 5.1 and I see no reason to move them until VMWare sorts out this mess. Version 6 was looking promising until these updates and patches started causing havok.
This bug only appears to affect Veeam if you're using VMWare quiescense. If you're not (and I'm guessing most people aren't because it's terrible) and just using normal AAP then it supposedly doesn't affect you.
That's pointed out in the link but is probably worth putting here. So for the most part we should all be fine.
3 of my hosts are on this patch level of 6.0 and 5 hosts are still on the latest patch level for 5.0.
I'm not updating my 5 other hosts until a new patch comes out (expected late july) that fixes ESXi 6.0 and vSphere replication 6.1 hostd crashes with "Panic: Assert Failed: "!repDiskInfo->GetDiskReplicationId().empty()" @ bora/vim/hostd/hbrsvc/ReplicationConfig.cpp:769" if you try to vMotion a replicated VM where only 1 (of many) disks are configured to replicate.
mcvosi wrote:Guess I was late to the party on this...
But yes, QA has really slipped at VMware these days. Every time I think I patch to a stable build something breaks with each new "fix."
Funny I had an argument with someone on IRC, the cisco channel I believe I was in at the time, (I'm in many of them), to which he kept saying ESXi6 is solid and suggested to move to it. Which I replied back with endless KB's from VMware on how terrible QA has been on it and that CBT in general has broke multiple times. The amount of links i had to reference has near endless... I asked for any references on its stability which I received none.