Cluster Re-balance Failure

[nimos001]

I have a custom datacenter defined to merge two clusters. My goal is to utilize this to help move workloads from one over burdened cluster to a brand new cluster with nearly no load. I am receiving recommendations as expected but when taking them within Operations Manager they all fail with the below message. I have verified that I can take these actions manually from the web client without any issues. I am not sure why this is failing trying to use Operations Manager. I checked over the policy to make sure there wasn't anything I needed to tweak but did not see anything. Does anyone have any ideas here? Thank you.

"The VM being moved does not share storage with all of the hosts on the destination cluster"

Environment Notes:

• vRealize Operation Manager 6.3
• vCenter 6U2
• ESXi 5.5U3d
• Storage is not shared across clusters

[carvaled]

the clusters (host's) you are trying to balance between must have shared vMotion vLans, VM Network, Storage and have CPU compatibility otherwise it will fail to move VM's in vCenter. (it will perform some checks before) ---> from the msg below it looks like the VM has a disk on a datastore that the destination host does not share...

I cant remember 100% but I think the "rebalance container" will only online vMotion between hosts and not between datastore's to balance the clusters...

Try to manually move in vCenter one of the VM's as suggested in 'rebalance container' and see what compatibility check messages you get.

Cheers

vMan

[nimos001]

You have been able to perform 'shared nothing' migrations since 5.1 if I am not mistaken? Also as mentioned I can perform these migrations manually without any errors or warnings.

[carvaled]

I assume you are trying to "Online" vMotion these machines, they are not powered off right?

The storage that the VM runs on must be available on the hosts they are moving to, I cant remember 100% but as far as I remember "Rebalance container" is not going to move your storage around just the CPU and MEM workload so the storage must be shared.

Dont forget to patch your vROPS if your are using automation... VMSA-2016-0016 - VMware Security & Compliance Blog - VMware Blogs

[nimos001]

I am attempting to migrate running workloads and as I mentioned this has been supported for a long time now. You do not have to have the same storage presented to both hosts. I can perform this action manually without any issues.

[carvaled]

i still think you have to have common datatstores for vRops to do the rebalancing... as far as i remember rebalance container is only looking at CPU and MEM workloads and will not move the storage (hence the failure)

VMware vRealize Operations Manager 6.2 Release Notes

Migration may fail if datastore is not accessible to network
VM migration across clusters may fail during a rebalance action with the message "Unable to access the virtual machine configuration: Unable to access file".

Workaround: Ensure that all hosts in every cluster participating in a rebalance action share the same datastores that every VM in every cluster participating in the rebalance action resides on.

Are both the clusters in the same datacenter in vCenter? or do the clusters belong to different datacenters?

[nimos001]

Same datacenter in vCenter...

[carvaled]

So looking at the release notes i pasted earlier and the error message you got... the storage must be shared for "rebalance container" to work..

Unfortunately i dont have an environment to test this on which have actions enabled on so i cant 100% confim it...

I raised the original requirement or MoveVM, custom datacenters and Rebalance Container to VMware and the early version didnt perform storage vMotions (it was about balancing out CPU and MEM contention)

[nimos001]

So, I should have also included this in my previous post, but while I am currently focused on balancing across '2' clusters, originally I had all 6 of our production clusters in this grouping and I was able to have it migrate across other clusters without shared storage. So that being said I do not believe not having shared storage across is the issue but am still digging into it. I am going to open up an SR and will post the result of that. I appreciate the replies.

[carvaled]

ok cool, well keep us posted as i would be interested with the outcome