cloudstack-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Tutkowski, Mike" <Mike.Tutkow...@netapp.com>
Subject Re: Cannot migrate VMware VM with root disk to host in different cluster (CloudStack 4.10)
Date Fri, 24 Mar 2017 00:03:43 GMT
I think I should open a blocker for this for 4.10. Perhaps one of our VMware people can take
a look. It sounds like it’s a critical issue.

On 3/23/17, 4:48 PM, "Tutkowski, Mike" <Mike.Tutkowski@netapp.com> wrote:

    OK, yeah, it does.
    
    The source host has access to the source datastore and the destination host has access
to the destination datastore.
    
    The source host does not have access to the destination datastore nor does the destination
host have access to the source datastore.
    
    I've been focusing on doing this with a source and a host datastore that are both either
NFS or iSCSI (but I think you should be able to go NFS to iSCSI or vice versa, as well).
    
    > On Mar 23, 2017, at 4:09 PM, Sergey Levitskiy <Sergey.Levitskiy@autodesk.com>
wrote:
    > 
    > It shouldn’t as long the destination host has access to the destination datastore.
    > 
    > On 3/23/17, 1:34 PM, "Tutkowski, Mike" <Mike.Tutkowski@netapp.com> wrote:
    > 
    >    So, in my case, both the source and target datastores are cluster-scoped primary
storage in CloudStack (not zone wide). Would that matter? For XenServer, that cluster-scoped
configuration (but using storage repositories, of course) works.
    > 
    >    On 3/23/17, 2:31 PM, "Sergey Levitskiy" <Sergey.Levitskiy@autodesk.com>
wrote:
    > 
    >        It looks like a bug. For vmware, moving root volume with migrateVolume with
livemigrate=true for zone-wide PS works just fine for us. In the background, it uses StoragevMotion.
From another angle MigrateVirtualMachine works also perfectly fine. I know for a fact that
vmware supports moving from host to host and storage to storage at the same time so it seems
to be a bug in migrateVirtualMachineWithVolume implementation. vSphere standard license is
enough for both regular and storage vMotion.
    > 
    >        On 3/23/17, 1:21 PM, "Tutkowski, Mike" <Mike.Tutkowski@netapp.com> wrote:
    > 
    >            Thanks, Simon
    > 
    >            I wonder if we support that in CloudStack.
    > 
    >            On 3/23/17, 2:18 PM, "Simon Weller" <sweller@ena.com> wrote:
    > 
    >                Mike,
    > 
    > 
    >                It is possible to do this on vcenter, but it requires a special license
I believe.
    > 
    > 
    >                Here's the info on it :
    > 
    >                https://pubs.vmware.com/vsphere-51/index.jsp#com.vmware.vsphere.vcenterhost.doc/GUID-A16BA123-403C-4D13-A581-DC4062E11165.html
    > 
    >                https://pubs.vmware.com/vsphere-51/index.jsp#com.vmware.vsphere.vcenterhost.doc/GUID-561681D9-6511-44DF-B169-F20E6CA94944.html
    > 
    > 
    >                - Si
    >                ________________________________
    >                From: Tutkowski, Mike <Mike.Tutkowski@netapp.com>
    >                Sent: Thursday, March 23, 2017 3:09 PM
    >                To: dev@cloudstack.apache.org
    >                Subject: Re: Cannot migrate VMware VM with root disk to host in different
cluster (CloudStack 4.10)
    > 
    >                This is interesting:
    > 
    >                If I shut the VM down and then migrate its root disk to storage in
the other cluster, then start up the VM, the VM gets started up correctly (running on the
new host using the other datastore).
    > 
    >                Perhaps you simply cannot live migrate a VM and its storage from one
cluster to another with VMware? This works for XenServer and I probably just assumed it would
work in VMware, but maybe it doesn’t?
    > 
    >                The reason I’m asking now is because I’m investigating the support
of cross-cluster migration of a VM that uses managed storage. This works for XenServer as
of 4.9 and I was looking to implement similar functionality for VMware.
    > 
    >                On 3/23/17, 2:01 PM, "Tutkowski, Mike" <Mike.Tutkowski@netapp.com>
wrote:
    > 
    >                    Another piece of info:
    > 
    >                    I tried this same VM + storage migration using NFS for both datastores
instead of iSCSI for both datastores and it failed with the same error message:
    > 
    >                    Required property datastore is missing from data object of type
VirtualMachineRelocateSpecDiskLocator
    > 
    >                    while parsing serialized DataObject of type vim.vm.RelocateSpec.DiskLocator
    >                    at line 1, column 326
    > 
    >                    while parsing property "disk" of static type ArrayOfVirtualMachineRelocateSpecDiskLocator
    > 
    >                    while parsing serialized DataObject of type vim.vm.RelocateSpec
    >                    at line 1, column 187
    > 
    >                    while parsing call information for method RelocateVM_Task
    >                    at line 1, column 110
    > 
    >                    while parsing SOAP body
    >                    at line 1, column 102
    > 
    >                    while parsing SOAP envelope
    >                    at line 1, column 38
    > 
    >                    while parsing HTTP request for method relocate
    >                    on object of type vim.VirtualMachine
    >                    at line 1, column 0
    > 
    >                    On 3/23/17, 12:33 PM, "Tutkowski, Mike" <Mike.Tutkowski@netapp.com>
wrote:
    > 
    >                        Slight typo:
    > 
    >                        Both ESXi hosts are version 5.5 and both clusters are within
the same VMware datastore.
    > 
    >                        Should be (datastore changed to datacenter):
    > 
    >                        Both ESXi hosts are version 5.5 and both clusters are within
the same VMware datacenter.
    > 
    >                        On 3/23/17, 12:31 PM, "Tutkowski, Mike" <Mike.Tutkowski@netapp.com>
wrote:
    > 
    >                            A little update here:
    > 
    >                            In the debugger, I made sure we asked for the correct
source datastore (I edited the UUID we were using for the source datastore).
    > 
    >                            When VirtualMachineMO.changeDatastore is later invoked
having the proper source and target datastores, I now see this error message:
    > 
    >                            Virtual disk 'Hard disk 1' is not accessible on the host:
Unable to access file [SIOC-1]
    > 
    >                            Both ESXi hosts are version 5.5 and both clusters are
within the same VMware datastore.
    > 
    >                            The source datastore and the target datastore are both
using iSCSI.
    > 
    >                            On 3/23/17, 11:53 AM, "Tutkowski, Mike" <Mike.Tutkowski@netapp.com>
wrote:
    > 
    >                                Also, in case it matters, both datastores are iSCSI
based.
    > 
    >> On Mar 23, 2017, at 11:52 AM, Tutkowski, Mike <Mike.Tutkowski@netapp.com>
wrote:
    >> 
    >> My version is 5.5 in both clusters.
    >> 
    >>> On Mar 23, 2017, at 9:48 AM, Sateesh Chodapuneedi <sateesh.chodapuneedi@accelerite.com>
wrote:
    >>> 
    >>> 
    >>>>> On 23/03/17, 7:21 PM, "Tutkowski, Mike" <Mike.Tutkowski@netapp.com>
wrote:
    >>> 
    >>>>> However, perhaps someone can clear this up for me:
    >>>>> With XenServer, we are able to migrate a VM and its volumes from
a host using a shared SR in one cluster to a host using a shared SR in another cluster even
though the source host can’t see the target SR.
    >>>>> Is the same thing possible with VMware or does the source host have
to be able to see the target datastore? If so, does that mean the target datastore has to
be zone-wide primary storage when using VMware to make this work?
    >>> Yes, Mike. But that’s the case with versions less than 5.1 only. In vSphere
5.1 and later, vMotion does not require environments with shared storage. This is useful for
performing cross-cluster migrations, when the target cluster machines might not have access
to the source cluster's storage.
    >>> BTW, what is the version of ESXi hosts in this setup?
    >>> 
    >>> Regards,
    >>> Sateesh,
    >>> CloudStack development,
    >>> Accelerite, CA-95054
    >>> 
    >>>  On 3/23/17, 7:47 AM, "Tutkowski, Mike" <Mike.Tutkowski@netapp.com>
wrote:
    >>> 
    >>>      This looks a little suspicious to me (in VmwareResource before we call
VirtualMachineMO.changeDatastore):
    >>> 
    >>>                      morDsAtTarget = HypervisorHostHelper.findDatastoreWithBackwardsCompatibility(tgtHyperHost,
filerTo.getUuid());
    >>>                      morDsAtSource = HypervisorHostHelper.findDatastoreWithBackwardsCompatibility(srcHyperHost,
filerTo.getUuid());
    >>>                      if (morDsAtTarget == null) {
    >>>                          String msg = "Unable to find the target datastore:
" + filerTo.getUuid() + " on target host: " + tgtHyperHost.getHyperHostName() + " to execute
MigrateWithStorageCommand";
    >>>                          s_logger.error(msg);
    >>>                          throw new Exception(msg);
    >>>                      }
    >>> 
    >>>      We use filerTo.getUuid() when trying to get a pointer to both the target
and source datastores. Since filerTo.getUuid() has the UUID for the target datastore, that
works for morDsAtTarget, but morDsAtSource ends up being null.
    >>> 
    >>>      For some reason, we only check if morDsAtTarget is null (I’m not sure
why we don’t check if morDsAtSource is null, too).
    >>> 
    >>>      On 3/23/17, 7:31 AM, "Tutkowski, Mike" <Mike.Tutkowski@netapp.com>
wrote:
    >>> 
    >>>          Hi,
    >>> 
    >>>          The CloudStack API that the GUI is invoking is migrateVirtualMachineWithVolume
(which is expected since I’m asking to migrate a VM from a host in one cluster to a host
in another cluster).
    >>> 
    >>>          A MigrateWithStorageCommand is sent to VmwareResource, which eventually
calls VirtualMachineMO.changeDatastore.
    >>> 
    >>>              public boolean changeDatastore(VirtualMachineRelocateSpec relocateSpec)
throws Exception {
    >>>                  ManagedObjectReference morTask = _context.getVimClient().getService().relocateVMTask(_mor,
relocateSpec, VirtualMachineMovePriority.DEFAULT_PRIORITY);
    >>>                  boolean result = _context.getVimClient().waitForTask(morTask);
    >>>                  if (result) {
    >>>                      _context.waitForTaskProgressDone(morTask);
    >>>                      return true;
    >>>                  } else {
    >>>                      s_logger.error("VMware RelocateVM_Task to change datastore
failed due to " + TaskMO.getTaskFailureInfo(_context, morTask));
    >>>                  }
    >>>                  return false;
    >>>              }
    >>> 
    >>>          The parameter, VirtualMachineRelocateSpec, looks like this:
    >>> 
    >>>          http://imgur.com/a/vtKcq (datastore-66 is the target datastore)
    >>> 
    >>>          The following error message is returned:
    >>> 
    >>>          Required property datastore is missing from data object of type
VirtualMachineRelocateSpecDiskLocator
    >>> 
    >>>          while parsing serialized DataObject of type vim.vm.RelocateSpec.DiskLocator
    >>>          at line 1, column 327
    >>> 
    >>>          while parsing property "disk" of static type ArrayOfVirtualMachineRelocateSpecDiskLocator
    >>> 
    >>>          while parsing serialized DataObject of type vim.vm.RelocateSpec
    >>>          at line 1, column 187
    >>> 
    >>>          while parsing call information for method RelocateVM_Task
    >>>          at line 1, column 110
    >>> 
    >>>          while parsing SOAP body
    >>>          at line 1, column 102
    >>> 
    >>>          while parsing SOAP envelope
    >>>          at line 1, column 38
    >>> 
    >>>          while parsing HTTP request for method relocate
    >>>          on object of type vim.VirtualMachine
    >>>          at line 1, column 0
    >>> 
    >>>          Thoughts?
    >>> 
    >>>          Thanks!
    >>>          Mike
    >>> 
    >>>          On 3/22/17, 11:50 PM, "Sergey Levitskiy" <Sergey.Levitskiy@autodesk.com>
wrote:
    >>> 
    >>> 
    >>>              Can you trace which API call being used and what parameters
were specified? migrateVirtualMachineWithVolumeAttempts vs migrateVirtualMachine
    >>> 
    >>> 
    >>> 
    >>> 
    >>> 
    >>> 
    >>> 
    >>> 
    >>> 
    >>> 
    >>> 
    >>> 
    >>> DISCLAIMER
    >>> ==========
    >>> This e-mail may contain privileged and confidential information which is
the property of Accelerite, a Persistent Systems business. It is intended only for the use
of the individual or entity to which it is addressed. If you are not the intended recipient,
you are not authorized to read, retain, copy, print, distribute or use this message. If you
have received this communication in error, please notify the sender and delete all copies
of this message. Accelerite, a Persistent Systems business does not accept any liability for
virus infected mails.
    > 
    > 
    > 
    > 
    > 
    > 
    > 
    > 
    > 
    > 
    > 
    > 
    > 
    > 
    > 
    > 
    

Mime
View raw message