Migration from 1.7 to 2.1

[whinpo]

Hi,

I'm working on a migration from 1.7 to 2.1.
I'm really impressed by the improvements I can see between the 2 versions!!! great job guyz!

I've got several questions :
1) I've mapped my old NFS storage used to host my .iso files. It connects on 2.1, I can see the space allocated and so on, but I see no files in it from the GUI. If I check in /mnt/pve/Iso, I can see it... why?
I've got the same issue on the NFS where i store my 1.7 backups...

2) I'm using an iSCSI storage on 1.7. Is it safe connect my 2.1 machines on the same LVM? I want to do that to allow the 2.1 servers to see the LVM, move the VM .conf file from the 1.7 servers to the 2.1 ones and start the VM... Is it possible? safe?

Thanks by advance for your help, remarks...

 

[tom]

Hi,

I'm working on a migration from 1.7 to 2.1.
I'm really impressed by the improvements I can see between the 2 versions!!! great job guyz!

I've got several questions :
1) I've mapped my old NFS storage used to host my .iso files. It connects on 2.1, I can see the space allocated and so on, but I see no files in it from the GUI. If I check in /mnt/pve/Iso, I can see it... why?
I've got the same issue on the NFS where i store my 1.7 backups...

check the new dir structure on 2.x. now we use only one share, and below we add several dir´s for ISO, backups, templates, ...

2) I'm using an iSCSI storage on 1.7. Is it safe connect my 2.1 machines on the same LVM? I want to do that to allow the 2.1 servers to see the LVM, move the VM .conf file from the 1.7 servers to the 2.1 ones and start the VM... Is it possible? safe?

Thanks by advance for your help, remarks...

be careful with this, do not access it the same time. I suggest you do the move in one run.

 

[whinpo]

Thanks Tom for your answers.

1) I'll check
2) When you tell me to be careful, you mean to be careful not to start the same machine twice, right? I manage few VM's (20), and I'll manually move the conf file of each machine (after having stopped it), so there should be no problem.

I was afraid of a new mechanism in the 2.1 that could wipe my actual LVM?

 

[NSW]

2) When you tell me to be careful, you mean to be careful not to start the same machine twice, right? I manage few VM's (20), and I'll manually move the conf file of each machine (after having stopped it), so there should be no problem.

I was afraid of a new mechanism in the 2.1 that could wipe my actual LVM?

What careful means in this situation is not having both the old version and new version accessing the same array at the same time. 2.1 wont wipe out your array unless you tell it to, but you can corrupt the data on it if both versions are accessing it at the same time.

I migrated my VM's from 1.9 to 2.1 by taking all the VM's offline, backing them up (safety in case there were problems), migrating everything to the master node for 1.9, disabling the iSCSI array for 1.9, removing the second node from the cluster and upgrading it to 2.1, attaching the iSCSI array to the new 2.1 server, pull back in my LVM groups/volumes and copying over the VM *.conf configuration files from the 1.9 server to the new 2.1 server.

You can get the full play by play on what i did here on my thread.

 

[m.ardito]

What careful means in this situation is not having both the old version and new version accessing the same array at the same time. 2.1 wont wipe out your array unless you tell it to, but you can corrupt the data on it if both versions are accessing it at the same time.

I migrated my VM's from 1.9 to 2.1 by taking all the VM's offline, backing them up (safety in case there were problems), migrating everything to the master node for 1.9, disabling the iSCSI array for 1.9, removing the second node from the cluster and upgrading it to 2.1, attaching the iSCSI array to the new 2.1 server, pull back in my LVM groups/volumes and copying over the VM *.conf configuration files from the 1.9 server to the new 2.1 server.

You can get the full play by play on what i did here on my thread.

i am in the middle of the conversion from 1.7 to 2.1
i backed up vm on 1.7, stopped but without delete them for now, and restored back them on 2.1 BUT changing their VMID ( i added 100 to VMID, so if vmid was 101 on 1.7 now it'd 201 on 2.1 and so on), because i'm using the same ISCSI space for vm disks. no problems until now, but i know it's risky. it's better use two different iscsi targets for old and new pve datacenters, since they know nothing about each other.

Marco

 

[m.ardito]

check the new dir structure on 2.x. now we use only one share, and below we add several dir´s for ISO, backups, templates, ....

is there any official info on this in the wiki? i couldn' find that... maybe i'm looking in the wrong places?

Marco

 

[whinpo]

Thanks for your answer.

But I can't proceed the way you propose (I first need to test the servers on 2.1...) and backup/restore would take too much time = downtime...

I wanted both version to access the same array, but of course not use the same VM (and the same LV) at the same time

I would like to proceed this way :
1) upgrade a server to 2.1 and attach it to iSCSI
2) move the conf file of a non-critical VM and run it under 2.1 (the VM would then only be seen on the 2.1 machine) and let it run for some days... to be sure everything is ok
3) if ok : move some other VM's on this server
4) migrate another server and test the cluster (moving VM is only what I need, no need for HA)
5) if ok : move the other VM's and migrate the servers

So a VM could only be seen either on the 1.9 cluster or on the 2.1 cluster, but there should no way to have it on both (and i'll of course double check). If I do it carefully there should be no risk of multi-access on the same Logical Volume.

Do you believe there is a risk anywhere? (and I repeat i'll take care not to run a VM twice)

regards

 

[NSW]

backup/restore would take too much time = downtime...

I didn't use the backup/restore method. I just took backups in case i needed them. What i did do was take the 1.9 server off the iSCSI array, mount the iSCSI array on the 2.1 server and pull in my original LVM groups/volumes with the original VM data on them. I copied just the small VM *.conf text files from 1.9 over to 2.1. I was able to start my original VM's with no problems. (outside of needing to redo the network card settings)

Do you believe there is a risk anywhere? (and I repeat i'll take care not to run a VM twice)

To be honest, i can't say one way or another. I'm not sure how an LVM manager running on one system will interact with another one. This was one reason i forced downtime to make the transition, i was not willing to risk corrupting a 16TB array with 10 VM's on it. The iSCSI array itself shouldn't care one way or another, it was designed to have multiple connections.

If you want to attempt it, make sure you have backups of everything. The biggest risk is having the entire array get corrupted, not just one VM.

 

[whinpo]

I indeed agree with "The biggest risk is having the entire array get corrupted, not just one VM."

But is this risk real?

Anybody else to enlight us?