upgrade path & second server to cluster

[m.ardito]

Hi, I need help to add a second ve server (as a cluster) and upgrade to 1.7

i am now running a single 1.5 (2.6.18) on IBM x3650M2, two nic card, four eth, since months. Works perfect, one openvz machine (local), a bunch of kvm on iscsi storage (qnap TS809RP), iso and backup storage on NFS.

proxmox1:~# pveversion -v
pve-manager: 1.5-10 (pve-manager/1.5/4822)
running kernel: 2.6.18-2-pve
proxmox-ve-2.6.18: 1.5-5
pve-kernel-2.6.18-2-pve: 2.6.18-5
qemu-server: 1.1-16
pve-firmware: 1.0-5
libpve-storage-perl: 1.0-13
vncterm: 0.9-2
vzctl: 3.0.23-1pve11
vzdump: 1.2-5
vzprocps: 2.0.11-1dso2
vzquota: 3.0.11-1
pve-qemu-kvm-2.6.18: 0.9.1-5

now, i have another identical server, it has been "empty" since today, when i installed 1.7 from scratch, just to chek that it ws installable with no problems, and it ended up quite well (i had just to piint put that eth0 and eth1 were assigned to the other nic card).

I am in doubt about which path to follow:
1) keep installed 1.7 on proxmox2, move all vm there (restoring backups), then upgrade proxmox1, then join both in a cluster

2) install 1.5 on proxmox 2, join both in cluster, then upgrade both to 1.7

3) ...? any other safer path, perhaps also minimizing downtime? what to double check to ensure nothing can go wrong?

need your help, pve comunity...

Marco

 

[tom]

1. install 1.7 on new server (go for the 2.6.18 branch).
2. upgrade existing one, shutdown all VM´s, reboot.
3. start VM´s, create cluster.

can´t you afford the downtime due to a reboot of the existing one?

 

[m.ardito]

nothing time critical but i wish to minimize downtime anyway.
i understand you suggestion, i supposed there could be a way to downtime just one vm at once, like:
1. install 1.7 on server2 (go for the 2.6.18 branch).
2. replicate on server2 the same configuration (system, storage, backup) as in first 1.5 server1
3. shut down one vm and restore it on the 1.7 server2, check everything, repeat for each vm.
4. upgrade (now empty) 1.5 server1 to 1.7 or reinstall from scratch
5. create cluster on server2 and join empty server1 to cluster

it's more complicated for sure, but do you feel it could be more a safe path...?
i like you simple way, but i fear upgrading a server that has all my vm and if it will not restart for some reason, i'll be stuck with all vm down?
i don't know, it's the first time i upgrade pve major version, and never tried a cluster, i just feel something will go wrong somewhere...

btw, why you suggest to go for 2.6.18, instead of newer 2.6.32?
is that a step to take later, once the cluster is stable, or not at all?

M

 

[tom]

pls note, you cannot reliable live migrate between different kernel versions and pve versionsy. could work but its untested.

I suggested 2.6.18 as you already using this one successfully - and if you use OpenVZ, its the stable branch here.

pve cluster: you should use the same version, means you cannot mix 1.5 with 1.7. but you can use different kernels on the nodes, but take care if you need live migration - identical versions works best.