Need Help - Setting up a 3 node VSan

[kb9jlo]

I evidently goofed something up when I was adding the third node (started by creating a 2 node v san from the "StarWind-Virtual-SAN-Hyper-Converged-2-Node-Cluster-VMware-vSphere" document).
Anyway I lost a 'device' on the original node (that I created the storage on originally). SEE ATTACHMENT FOR SCREENSHOT

I tried adding a device, existing, new - that didn't work. Couldn't add it either way.

The drive shows up in VMware, has data on it. But now I'm afraid to use it.

[tig_jeff]

Do you get an error that you cannot create the device because it exists?

[Vitaliy (Staff)]

Hello kb9jlo,

The drive shows up in VMware, has data on it. But now I'm afraid to use it.

Could you please check whether in VMware storage is available from 10.1.155.15 host?
If not, you can recreate a replica from the partner host to current host and your device should be avaialble from both.

[kb9jlo]

Sorry I never replied -- I didn't get an email - didn't know anyone responded.
I should have kept on checking. See my two replies.

[kb9jlo]

Do you get an error that you cannot create the device because it exists?

I did get an error - don't remember what it was. I was going through the motions of trying to recreate the target today but was afraid to push the 'finish' button -- not knowing what effect on a live system it would have.

There is so much I DON'T know that I don't know where to hesitate or where to jump forward...

I'm going to make another post as well. I need a "care and feeding of Starwind" manual.

[kb9jlo]

Hello kb9jlo,

The drive shows up in VMware, has data on it. But now I'm afraid to use it.

Could you please check whether in VMware storage is available from 10.1.155.15 host?
If not, you can recreate a replica from the partner host to current host and your device should be avaialble from both.

My 3 nodes is in production. I was against the wall and had to push forward. It's working but I'm at a disadvantage not knowing what it is doing. A week ago one host spontaneously rebooted (unrelated event). and the whole system worked. VM HA failed everything over and nothing bad happened.
It appears the data is there - the target in the SW console is just "gone". The file is actually on the disk. VMware "thinks" it's there and there are VM's using that datastore.

Completely unrelated one of my Starwind nodes (SW1) stopped working yesterday. I was able to restart the service and everything appears to be trying to resync back.

I have 3 nodes, SW1-3 and three datastores DS-1 thru 3. Each is just below 6 TB. The sync network is 10 Gb infiniband, each node has (2) dual port cards. All the nodes are daisy chained together. Alex Khorolets helped me through email through some spots but he's so busy he never had time to email me.

[Sergey (staff)]

You can connect StarWind image file to the existing target Storage1, If it is available on disk. Try this: Add Device(Advanced) --> Virtual Disk --> Use an Existing Virtual Disk --> Attach to Existing Target --> Create. The other alternative would be replicating this image file from the other node. The Datastore is available in vSphere Client which means there is an active path from the other node, so you can easily replicate.

[kb9jlo]

You can connect StarWind image file to the existing target Storage1, If it is available on disk. Try this: Add Device(Advanced) --> Virtual Disk --> Use an Existing Virtual Disk --> Attach to Existing Target --> Create. The other alternative would be replicating this image file from the other node. The Datastore is available in vSphere Client which means there is an active path from the other node, so you can easily replicate.

Thanks. I think what I'm going to have to do is delete the old "target" file (both disk & flash) and then recreate it or as you said sync it from another node.

I followed the 2 node VMware & 3 node VMware instructions you guys have posted -- and I think at one point I deleted the target's "connection" But since the file already exists now, and there is no free space that's why I can't "recreate" it.

[Boris (staff)]

kb9jloб
The procedure you described looks reasonable. Let us know if you need more information.

[kb9jlo]

It seems to have worked. Disks are syncing...

[kb9jlo]

kb9jloб
The procedure you described looks reasonable. Let us know if you need more information.

I have a problem - when I setup the new replica it didn't ask me for the cache memory and set it at 1 MB. I'm experiencing high latency numbers now. I'm assuming because of this. And now this new replica is the master for that particular disk/LUN...
I've tried the Cache Setting (experimental) to reset it but it does nothing at all.
How can I correct this?

[Sergey (staff)]

You can change cache setting, by editing StarWind devices header files. I hope this will help:
https://knowledgebase.starwindsoftware. ... -l1-cache/
Make sure to backup original headers, stop StarWindService before changing headers and do this procedure on both nodes.

[kb9jlo]

You can change cache setting, by editing StarWind devices header files. I hope this will help:
https://knowledgebase.starwindsoftware. ... -l1-cache/
Make sure to backup original headers, stop StarWindService before changing headers and do this procedure on both nodes.

If the node I do this on is the "master" what happens exactly? Is there an easy way to "move" the master?

[Boris (staff)]

If the node I do this on is the "master" what happens exactly? Is there an easy way to "move" the master?

Could you please explain more what you meant by "master" here?
If "master" meant node priority, it has no influence on the operations you need to perform.
In fact, what is important before you do any changes to the header file is a couple of steps. They include checking whether StarWind devices are synchronized on both nodes and making backups of both .swdsk files - the ordinary .swdsk and *_HA.swdsk (just in case, when you spoil anything during editing, you will have a working set of files to revert to upon service restart).

[kb9jlo]

If the node I do this on is the "master" what happens exactly? Is there an easy way to "move" the master?

Could you please explain more what you meant by "master" here?
If "master" meant node priority, it has no influence on the operations you need to perform.
In fact, what is important before you do any changes to the header file is a couple of steps. They include checking whether StarWind devices are synchronized on both nodes and making backups of both .swdsk files - the ordinary .swdsk and *_HA.swdsk (just in case, when you spoil anything during editing, you will have a working set of files to revert to upon service restart).

Sorry, not sure what terminology to use. And I probably move the "master" on the VMware side. I will explore more there.
When I say "master" I mean the main node in the VMware HA cluster. See the screenshot.