Cloudstack 4.20.0.0, KVM with CEPH RBD live migration error

[sourceYan]

Hi cloudstack Community,

we do some tests with cloudstack and ceph.
Create the primary ceph storage is no problem, offline migration works also good!
But if i want do move a VM from or to ceph, i got the following error:

Failed to migrate VM [VM instance {"id":10,"instanceName":"i-2-10-VM","type":"User","uuid":"ff8ae527-bd26-4b74-b02c-8de21159ff01"}] along with its volumes due to [com.cloud.utils.exception.CloudRuntimeException: Copy volume(s) to storage(s) [{volume: "17", from: "7", to:"9"}] and VM to host [{vm: "10", from: "1", to:"4"}] failed in StorageSystemDataMotionStrategy.copyAsync. Error message: [Exception during migrate: org.libvirt.LibvirtException: Storage pool not found: no storage pool with matching target path '/var/log/cloudstack/management/cloudstack'].].

any ideas?

best regards!
YAN

[DaanHoogland]

@sourceYan , migrations across different storage solutions don’t always work great. It might well be that yours isn’ t supported.
What is the target storage for type (i.e. NFS)?
Did you try the reverse, to your ceph cluster?

[sourceYan]

hi @DaanHoogland - thank you for your message!
i tested both, and reverse NFS Storage an "local" Mountpoint (OCFS2, iSCSI)... but all fail.
Only offline it's possible. Between OCFS and NFS it works without issues.

[weizhouapache]

this is a known limitation I think

for vm live mirgation between ceph storages, we should migrate the volume instead of the vm (it will use "virsh blockcopy").
but it may not with either.

[sourceYan]

hi @weizhouapache - thanks for your message!
if i try to migrate only the storage, a message appear, that with kvm i have to move also the host.
i didn't tested the migration between two ceph storages yet...

[DaanHoogland]

@sourceYan , did you find a way forward or a workaround?

I think there is no life migration btween Ceph and NFS. (As I understand this is what you want !?)

[sourceYan]

@DaanHoogland, i though live migration between different storage types is possible, but that's not.
That was not clear for me, if it's possible or not...
from the technical perspective, it should be possible, if the host has both storages accessible/mounted?

[DaanHoogland]

for each combination of storages this needs to be implemented separately. Not many of them are implemented now. BTW all of them would be pseudo life, as the backing disk would have to switch mid-operation and the VM would always have to pause for that.

[weizhouapache]

+1 with @DaanHoogland

@sourceYan
to be clear, do you want live-migrate a vm between ceph storage pools ?

[sourceYan]

no, initially i want to migrate from nfs to ceph.
Currently we rebuild the lab, if we're up again, i test the different migrations (from nfs to ceph, ceph to ceph...)

[DaanHoogland]

@sourceYan , do you have a way forward by now?

[sourceYan]

we're currently rebuilding the lab. I'll post my results here when we're back.