CDP, Maintenance Mode and Inaccessible Objects

[lwarner]

I have CDP Policies that are running perfectly. Both source and target sites use VSAN. I went to do maintenance at the Target site, and it took *hours* for each host to enter Maintenance mode using Ensure Availability.

Since my Target site is not very "hot" I ran a Data Migration Pre-check to see what would happen if I vMotioned all VMs off each host (including the powered off replica VMs) and used "No Data Migration" for Maintenance mode... it showed a whole bunch of Inaccessible Objects.

So then I looked at the VSAN Virtual Objects, and it shows lots of Inaccessible Objects, but only for VMs that are in my CDP policies. The number of objects changes over time, some VMs will gain objects, some will lose them.

Questions. 1, are Inaccessible Objects part of the way CDP works? Can I remove them, as I would normal Inaccessible Objects? How do I make the hosts enter Maintenance Mode faster?

[Mildur]

Hi Leigh

I suggest to open a support case and let our support check this objects before you start removing them.

How do I make the hosts enter Maintenance Mode faster?

I assume it's about maintenance mode of a ESXi host. Maintenance Mode for a host will be activated after all VMs are moved or shutdown.
I would start analyzing the performance of enabling maintenance mode on the ESXi host itself.

Best,
Fabian

[lwarner]

Just for info regarding Inaccessible Objects in case anyone else sees this, I opened a support ticket, 05841095, and although the exact cause was not discovered, we found that the issue appears to be resolved in v12.

[lwarner]

Another updates, it appears now that the issue persists into v12. Support is investigating.

[veremin]

The mentioned ticket was closed last week. Am I correct that the investigation continues under the new case (05865574)? I'm asking so we can follow it internally. Thanks!

[lwarner]

Vladimir, the Inaccessible Objects issue turned out to be with VMware. Some VSAN metadata does not update in real time; it seems to refresh perhaps every 30 - 90 minutes (just a wild guess based on what I see). Veeam objects appear to be dynamically created and removed at a moderately fast rate with CDP (normal). When you go to the VMware Cluster > Monitor > Virtual Objects, that information is also updated fairly frequently. However, the data driving the Virtual Objects (VSAN metadata) is always many minutes old, so it is displayed as Inaccessible. If you go to the VSAN Datastore view and hit Refresh - no more Inaccessible objects. So the short of it is, it's a cosmetic issue.

Regarding the unreasonable time to enter maintenance mode at the target site: after upgrading from v11 to v12, a ton of CDP issues went away, including this issue. CDP in v12 has been doing well!

[veremin]

Thank you for additional information and your kind words. I've shared the feedback with the R&D team and they deeply appreciate it.

[gschaffer]

The answer is simple. tlog files created by veeam has FTT=0 which means, they have only copy on one server. if the server fails, you lost your replica.
Obviosly support will investigate and will never find the issue. You need the developer to solve this.
v12.1 remains the same behavior.