pve cluster live migration setup

[screenie]

hi,

when i understand it right - for a pve cluster with live migration i need a drbd active/active+gfs2 or a shared mounted nfs on both nodes where the vm's are stored, correct?

or is there a additional way for doing this? without shared storage?

and what is the recommendation for such setup?

thx

 

[udo]

hi,

when i understand it right - for a pve cluster with live migration i need a drbd active/active+gfs2 or a shared mounted nfs on both nodes where the vm's are stored, correct?

or is there a additional way for doing this? without shared storage?

and what is the recommendation for such setup?

thx

Hi,
for openvz you need only local storage and kernel 2.6.18.
For kvm you need shared storage. DRBD is "shared" storage at two nodes... but you don't need gfs2. Lvm do the trick - on the drbd-storage is a volumegroup and the VM-disks ar logical volumes which are only accessed from one node at a time.

Recommendations? It's depends... speed, cheap,... i have good experience with a FC-SAN and LVM-storage on the LUNs.

Udo

 

[screenie]

ok, drbd without the whole rhcs makes it much easier
i am right that you are using a separate lvol for each vm?
one lvol for all my vm's are enough because when the backup job run's all vm's can be processed with the same snapshot - or is there something i missed?

so, i will try the drbd variant - it's cheap and i don't expect excessive disk-io's...

thx

 

[udo]

ok, drbd without the whole rhcs makes it much easier
i am right that you are using a separate lvol for each vm?
...

Hi,
pve will do that for you - if you create a disk for a VM it's create a logical volume:

# lvdisplay | grep dev/drbd
  LV Name                /dev/drbdvg/vm-103-disk-1
  LV Name                /dev/drbdvg/vm-111-disk-1
  LV Name                /dev/drbdvg/vm-126-disk-1
  LV Name                /dev/drbdvg/vm-121-disk-1
  LV Name                /dev/drbdvg/vm-132-disk-1

It must be independent LVs, otherwise the migration can't work.

Udo

 

[screenie]

ah ok - didn't know that...thx for clarifying...