Proxmox 2.1 dilemma (from 1 Host to 2)

[andrea68]

Hi,

I have 1 Proxmox 2.1 host (is an old 1.5 upgraded step-by-step to 2.1) with 2 VM on top (200GB each, 4GB ram) on local storage.

Now I have a second host configured nearly identical (its' a double Xeon, 6 SAS 10K rpm * 300GB configured in raid5, 24 Gb RAM, Power Redundancy).

But how can I configure this two hardware in Cluster+HA+drbd? Without destroy or delete my VM's?
This two VM contain services and I cannot destroy them.
I can just stop them for a couple of hours...

So if I want to build a drbd scenario with 2 physical host I must reformat disk drives, it's correct?
But I must also reinstall Proxmox?

Maybe can I use this strategy:

1. Setup on host2 with DRBD
2. Backup VM from host1 > host2
3. Restore VM on host2
4. Shutdown host1
5. Reconfiguration host1 and resync with host2

It's correct?
Any thought about it?

Thanks in advance.

 

[udo]

Hi,

I have 1 Proxmox 2.1 host (is an old 1.5 upgraded step-by-step to 2.1) with 2 VM on top (200GB each, 4GB ram) on local storage.

Now I have a second host configured nearly identical (its' a double Xeon, 6 SAS 10K rpm * 300GB configured in raid5, 24 Gb RAM, Power Redundancy).

But how can I configure this two hardware in Cluster+HA+drbd? Without destroy or delete my VM's?
This two VM contain services and I cannot destroy them.
I can just stop them for a couple of hours...

So if I want to build a drbd scenario with 2 physical host I must reformat disk drives, it's correct?

Hi Andrea,
yes that's correct

But I must also reinstall Proxmox?

No. Only all nodes which join the cluster must be empty.

Maybe can I use this strategy:

1. Setup on host2 with DRBD

I would do as first step backup all VMs from node2 and restore them to node1 (and start the VM on node1). After that create two (for each server one) drbd-volumes with lvm-ontop on node2, Create an cluster on node1 and join the empty node2 to the cluster. Define the lvm-drbd-storage as storage (available only on node2)

2. Backup VM from host1 > host2
3. Restore VM on host2

right - and restore on the right drbd-volume (depends on which node the VM should running later.

4. Shutdown host1
5. Reconfiguration host1 and resync with host2.
...

Change the drbd-storage after that (available on both nodes) and use live-migration to migrate the VMs from host2 to host1.

To make the live easier, you should use a simple server as third node. With only 2 nodes you run in trouble if one node died (quorum) esp. with HA!

Udo

 

[andrea68]

Hi Andrea,

[snip]

To make the live easier, you should use a simple server as third node. With only 2 nodes you run in trouble if one node died (quorum) esp. with HA!

Udo

Thanks for your hints.

My goal is to use at least 2 nodes and avoid that if a node dies, VM's also dies.
Do you think that without a third node it's impossible to obtain this?

This guide is still valid?
http://pve.proxmox.com/wiki/Two-Node_High_Availability_Cluster

[I don't need a "real" automatic live migration, it will be sufficient that if a node die I can start even manually my VM's...]

...

Thanks.

 

[e100]

If you want HA use three nodes, the third node could be some tiny Intel atom server if space/cost is an issue.

If you just want your data stored redundant and only want two servers use DRBD but do not configure HA VMs in Proxmox.

Two old posts of mine that explain why three nodes is better than two:
http://forum.proxmox.com/threads/8693-is-there-an-alternative-to-fencing-for-HA?p=49323#post49323
http://forum.proxmox.com/threads/8693-is-there-an-alternative-to-fencing-for-HA?p=49835#post49835

One Proxmox cluster I manage has 17 nodes, 14 of those nodes use DRBD (7 pairs)
We have fencing and failover domains configured ready to setup HA VMs.
So far we have decided not to configure any HA VMs for two reasons:
1. We do not need a zillion 9's of uptime
2. I am still concerned with using HA with DRBD, why am I concerned?:
http://forum.proxmox.com/threads/9121-High-Availability-2?p=51716#post51716

 

[andrea68]

If you want HA use three nodes, the third node could be some tiny Intel atom server if space/cost is an issue.

Thanks, this is interesting....

If you just want your data stored redundant and only want two servers use DRBD but do not configure HA VMs in Proxmox..

Ok! Maybe this is a right strategy.
As I said, maybe HA is too much for my needs: I just looking for a solution that save my VM if a failure occured to one server, so the other one can restart the VM.

Two old posts of mine that explain why three nodes is better than two:
http://forum.proxmox.com/threads/8693-is-there-an-alternative-to-fencing-for-HA?p=49323#post49323
http://forum.proxmox.com/threads/8693-is-there-an-alternative-to-fencing-for-HA?p=49835#post49835

Thank you very much, I will read this soon...!!!

One Proxmox cluster I manage has 17 nodes, 14 of those nodes use DRBD (7 pairs)
We have fencing and failover domains configured ready to setup HA VMs.
So far we have decided not to configure any HA VMs for two reasons:
1. We do not need a zillion 9's of uptime
2. I am still concerned with using HA with DRBD, why am I concerned?:
http://forum.proxmox.com/threads/9121-High-Availability-2?p=51716#post51716

Thanks again for your explanations!

 

[andrea68]

Hi Udo, I'm returning to this question, just for some clarification:

1 - For now I have only 2 physical host but in my case I don't need "real HA" I just need the possibility to migrate manually the VM's if a node fail, so I need a storage seen by 2 host.
So, if I don't have a SAN or iSCSI, my only choice is a DRBD storage from Node1+Node2 Drives.
It's correct?
Later if I add a third node can I use also HA, right?

2 - If previous assumptions are correct, how can I built DRBD without wipeout my first node that contain in local storage my VM's?
I'm not sure I understant correctly your previous hints and I dont want screw up everything

Thanks in advance...

 

[andrea68]

Any news? Thanks...

 

[udo]

Any news? Thanks...

Hi,
like I wrote before. If all VMs on node 1 (node 2 is empty) create an cluster on node1 (have an backup of all VMs of course). Join the empty node 2.
Create on node 2 two drbd-resources and bring them up, create two lvms ontop and define this as storage in the GUI (shared, available only on node 2).
Now you can stop a vm, create an backup and restore tis on node 2 with storage on the drbd-resource.
If you move all VMs from node 1 to node 2 you can create the same drbd-resorces on node 1 and do an sync from node 2 to node 1. After syncing switch to primary/primary, activate both lvms on node1, mark the storage as accessible for both nodes and use live-migration to move the VMs back to node 1.

Perhaps you should play with two test-systems before to see what happens.

Udo

 

[andrea68]

Thank you again, I will try.

Do you think it's possible virtualize proxmox to test this?
Thanks!