iscsi related

[thesubmitter]

Hi,

So just wanted to be clear on a few items

1. You don't share LUNs between Nodes (the shared option means what then - for images and non-dynamic data?)

2. Is the Recommended method is LVM on top of an ISCSI LUN per Physical Node. What about 1 LUN per VM, so you have many LUNs per physical node?

3. For Live Migration, this is possible but in reality what is happening is a copy on the SAN from 1 LUN to the other, not just pointing the new machine at the old location.

4. There is no multipath support, which begs the question that why not just mount the ISCSI/multipath LUN as a regular mount point and treat it as a directory.

5. More practically speaking, after I added the ISCSI target, then I added the LVM Group - I never get it offered as a storage backend for a container. I notice it says "content: images" while the local storage says "images, isos, CTs etc.....". Not sure where that is configured.

Thanks for ProxMox, great product, I've used it in another environment, I think its been running for almost 2 years completely undisturbed. Evaluating migrating a XEN w/ SAN-ISCSI environment to proxmox and running some tests/benchmarks.

Thanks!

 

[dietmar]

1. You don't share LUNs between Nodes (the shared option means what then - for images and non-dynamic data?)

What ? (we share iSCSI LUNs).

2. Is the Recommended method is LVM on top of an ISCSI LUN per Physical Node. What about 1 LUN per VM, so you have many LUNs per physical node?

You can use both setups - but LVM is more flexible.

3. For Live Migration, this is possible but in reality what is happening is a copy on the SAN from 1 LUN to the other, not just pointing the new machine at the old location.

Someone is confused, either you or me.

4. There is no multipath support, which begs the question that why not just mount the ISCSI/multipath LUN as a regular mount point and treat it as a directory.

There is multipath support (install multipath-tools). But I don't really get that question. Mounting a file system does not work with shared storage devices.

5. More practically speaking, after I added the ISCSI target, then I added the LVM Group - I never get it offered as a storage backend for a container. I notice it says "content: images" while the local storage says "images, isos, CTs etc.....". Not sure where that is configured.

Containers need a file system (NFS, local storage).

 

[thesubmitter]

Thanks for you responses

I assume this is still true: http://forum.proxmox.com/threads/9257-OpenVZ-and-ISCSI
thus openVZ only works on local storage?


Also, all my questions below were regarding openvz containers and I guess all your responses were regarding KVM VMs?

______

If that is the case can I just bind mount LUNs to Local Folders and add add them as local storage (after I create a FS on them) and if I do that, then I will definitely need 1 LUN per Container to avoid corruption right?

Thanks

 

[thesubmitter]

I thought v2 was going to add openvz on iscsi, oh well.

So now the question what is the best way to have redundancy and/or HA + migration and/or live migration with openVZ containers?

What I am thinking is:

1. I create many LUNs - 1 per Container that way I can reallocate the LUNs to any node.

2. In the event of a migration or live migration, is there a way to tell another node to start an already existing container on a given LUN that was previously mounted locally on a failed server or will I just be doing a standard migration (i.e: copying files between the LUNs?)

Thanks in advance

 

[dietmar]

You can use NFS for OpenVZ if you do not want to copy files on live migration.