Cluster Storage configuration help request.

[Astorian]

Hello,
i am fairly new to the Proxmox VE, thus i am looking for help here. We build up a node and after testing we took it in use.
Later, happy with the node we build an HW identical node and both two subscriptions. Updated both nodes 6.2-15 48bd51b6
I have configured the second node with identical storage directory like on first node - /dev/sdb mounted as /mnt/pve/VM_DATA
Then configured the cluster and realized the VM_DATA is not shown on the Host2
Reading the documentation, i found out that Directory storage cannot be shared.
There are VMs running on first node. What would be a good way to create a shared storage and migrate the VMS in this case?
regards, Astorian

 

[wolfgang]

Hi,

You can make a storage migration, but this does not mean that the storages are sync.
If you like replication, you have to use ZFS.
With ZFS, Proxmox VE provides storage replication [1].

1.) https://pve.proxmox.com/wiki/Storage_Replication

 

[Astorian]

Hi Wolfgang,
[i kann auch Deutsch wenn du so weiter machen willst ]
Thank you for responding. It seems i have not explained my self correctly.
I wanted to be able to migrate a VM to the second host only when needed. For example for host maintenance and so on.
I misunderstood and presumed that identical host storage config in cluster, will enable me to move the vm data with migration from one host to another.
I was able to add the directory from second host to the cluster but not with the same name VM_DATA as it already exists. VM_DATA1 is now there. Here is a screenshot:

There seems to be no possibility to select different storage in Migration. Am i only able to manually copy?

Yes, replication makes more sense if we want to use the cluster, as it is easier to migrate when the data is already synced.
Should i remove the VM_DATA and create a ZFS pool on both hosts, or can i do it one by one? So that i can move VMs to host with ZFS, redo the other host with ZFS and then finaly replicate?
The VM_DATA is a RAID10 volume on 4disks created on a hardware raid with 2GB cache - i presume i would have to remove raid and make the zfspool on the four disks.

 

[wolfgang]

The VM_DATA is a RAID10 volume on 4disks created on a hardware raid with 2GB cache - i presume i would have to remove raid and make the zfspool on the four disks.

Then ZFS is not an option for you.

But if it is ok for you to just migrate when both hosts are running you can do this.
The problem with your setup is that you named the datastores differently.

When the datastore naming is the same you can migrate on the GUI.
If they are different you must use the command line.

qm migrate .... --targetstorage <storage name>

You can set the node scope on the storage.
Then you can migrate the VM/CT to the shared storage.

 

[Astorian]

The problem with your setup is that you named the datastores differently.

When the datastore naming is the same you can migrate on the GUI.

Hi Wolfgang, thank you for your reply.
Before i created the cluster i have named both directories the same name:


However, after creating the cluster, only the host1 VM_DATA was shown. I wanted ot add it with the same name but i could not since the name was already defined:

Then i read in the documentation that directory cannot be shared. Is that why it is not using the directories as a cluster datastore?

 

[Astorian]

Could please someone tell me what would be the best way to redo my config or the whole cluster or point me to the documentation?
I would really like to have both storages accesible from both nodes.

 

[dcsapak]

if you want local disks to be shared across nodes, you have to have some kind of cluster storage, e.g. ceph/gluster or something like ocfs2 or gfs2