CSV setup with existing data

[Ardie]

Hello,

Initially I did not have access to my second host, so when I install and configured VSAN I connected to the local iSCSI and made them online and assigned drive letters (Y, Z).

I now have VM's stored in these Y/Z drives and now have the second host online. I configured VSAN replication and waited for it to complete which I now believe I should not have done yet.

I am now basically at step 79 in the guide:
https://www.starwindsoftware.com/resour ... erver-2016

I assume the CSV setup is going to want the volumes to be empty.

As I already have data on the volume, and drive letters assigned - what is the best way to complete these steps? Can I create the CSV with only the second host, migrate VM's to the new CSV, and then add the first host again after?

Thanks for your advice.

[Boris (staff)]

Bring the disks online on the second node and the node will recognize partitions and data on them.
If you have any other GUI-enabled Windows OS there, just connect to the second node using the Disk Management snap-in and you will be able to see the disks and bring them online. Otherwise, you can do that via PowerShell, I believe. Something like:

Code: Select all

Get-Disk | Where-Object IsOffline –Eq $True | Set-Disk –IsOffline $False

[Ardie]

Hi Boris,

I can bring the disks online - that shouldn't be a problem. The issue will be when I then start the setup for CSV - I believe CSV requires the disks to be blank. Is this incorrect?

Thanks.

[Boris (staff)]

No, It does not. CSVFS is introduced on top of NTFS. When both hosts are members of the cluster, both of them will be able to recognize that file system. If the second node is not yet there, it will display the CSVFS disks as RAW, but not CSVFS.

[Ardie]

Following along with the same guide, I started the verification tests (step 83) and almost immediately all of the iSCSI targets became inaccessible and of course all VM's had an improper shutdown. Currently I have multiple VM's running on two different VSAN HA images, they are all currently running on "node 1" and there is nothing on "node 2" right now.

If I refreshed the Targets in iscsicpl they were still all showing connected, however in diskpart the disks were no longer visible. After restarting the "starwindservice" service, it took a few minutes but the disks returned but were all marked offline. I made them online and also had to manually remove the readonly flag from each volume. I was then able to start my VMs again.

I assume the cluster verification tests caused a network disconnect, but as my iSCSI target is 127.0.0.1 I assume this should not cause any issues. Can you provide any advice on what to look into as to why this happened?

Additionally, when I check iscsicpl now on node1 where all my VMs are running, I do not see any targets listed at all for some reason from either host/node. The disks are however still connected and everything is functioning normally.

I am certainly interested in getting this product working and I know attempting this in production is not ideal, but I am not sure where to go from here.

Thanks.

[Oleg(staff)]

During the process of validation Cluster is checking the connections to CSVs and simulate their disconnections, even simultaneous disconnections on both sides. It could be the possible reason for the VMs fail.
Also, after adding to the Cluster, CSV is reserved by CSVFS.

I am certainly interested in getting this product working and I know attempting this in production is not ideal, but I am not sure where to go from here.

That is the different levels of interaction. Please check the Cluster validation report.

[Ardie]

The cluster verification was being done from a virtualized server, and as the test for some reason killed the iSCSI connection to VSAN, the VM powered off and verification did not complete so I have no logs to look at.

I believe I will need to attach temporary storage and migrate VHD's back from the iSCSI volumes before continuing unless you have any other suggestions. Unfortunately I cannot have production VMs getting shut down improperly during cluster verification.

[Oleg(staff)]

During the validation process Cluster is simulating disconnections of this ISCSI connections. As I told, even simultaneous disconnections on both hosts.
It is better to shut down the VMs or move to another storage in order to avoid the situation you had.