Dear David, Prasanna and all,
Good day to you, and thank you for your e-mail.
Thanks for the information and confirmation that the cloudstack-agent runs
as root and that CloudStack uses qemu+tcp (instead of qemu+ssh) on live
migration. Have also confirmed that auth_tcp is configured as none on
libvirtd on both hosts.
/etc/libvirt/libvirtd.conf:
auth_tcp="none"
I have tried simulating a live migration directly on the hypervisor host
itself, and I can confirm that I can do the live migration manually (not
through CloudStack).
On the first host (hv-kvm-01):
root@hv-kvm-01:~# virsh list --all
Id Name State
----------------------------------------------------
2 s-299-VM running
3 v-261-VM running
4 i-2-273-VM running
5 i-2-281-VM running
* 6 i-2-278-VM running*
7 i-33-287-VM running
root@hv-kvm-01:~# virsh migrate --live i-2-278-VM qemu+tcp://
root@10.237.3.22/system
After the migration:
root@hv-kvm-01:~# virsh list --all
Id Name State
----------------------------------------------------
2 s-299-VM running
3 v-261-VM running
4 i-2-273-VM running
5 i-2-281-VM running
7 i-33-287-VM running
On the second (target) host (hv-kvm-02):
root@hv-kvm-02:~# virsh list --all
Id Name State
----------------------------------------------------
2 r-271-VM running
3 i-2-275-VM running
4 i-2-276-VM running
5 i-34-289-VM running
6 i-2-293-VM running
* 8 i-2-278-VM running*
I can also do a live migration back from hv-kvm-02 to hv-kvm-01.
However, live migration through CloudStack GUI is still failing with the
same error message.
===
2013-07-24 11:47:36,944 DEBUG [agent.transport.Request]
(Job-Executor-1:job-658) Seq 32-1333264394: Sending { Cmd , MgmtId:
161342671900, via: 32, Ver: v1, Flags: 1001
11,
[{"MigrateCommand":{"vmName":"i-2-281-VM","destIp":"10.237.3.22","hostGuid":"00dd5dba-7419-3689-bc51-1671035c0d8f-LibvirtComputingResource","isWindows":false,"wait"
:0}}] }
2013-07-24 11:47:37,712 DEBUG [agent.transport.Request]
(AgentManager-Handler-3:null) Seq 32-1333264394: Processing: { Ans: ,
MgmtId: 161342671900, via: 32, Ver: v1, F
lags: 110, [{"MigrateAnswer":{"result":false,"details":"End of file while
reading data: Input/output error","wait":0}}] }
2013-07-24 11:47:37,713 DEBUG [agent.manager.AgentAttache]
(AgentManager-Handler-3:null) Seq 32-1333264394: No more commands found
2013-07-24 11:47:37,713 DEBUG [agent.transport.Request]
(Job-Executor-1:job-658) Seq 32-1333264394: Received: { Ans: , MgmtId:
161342671900, via: 32, Ver: v1, Flags: 1
10, { MigrateAnswer } }
2013-07-24 11:47:37,713 ERROR [cloud.vm.VirtualMachineManagerImpl]
(Job-Executor-1:job-658) Unable to migrate due to End of file while reading
data: Input/output error
2013-07-24 11:47:37,713 INFO [cloud.vm.VirtualMachineManagerImpl]
(Job-Executor-1:job-658) Migration was unsuccessful. Cleaning up:
VM[User|CentOS59-64bit]
===
Therefore, it seems that the problem is on the CloudStack's end rather than
libvirt's and/or QEMU/KVM's end. I am really at loss on how to resolve this
problem now.
Any help is greatly appreciated. Live migration is very important and I
can't move forward without this important feature.
Looking forward to your reply, thank you.
Cheers.
On Wed, Jul 24, 2013 at 12:27 AM, Prasanna Santhanam <t...@apache.org> wrote:
> On Tue, Jul 23, 2013 at 08:18:31AM -0600, Marcus Sorensen wrote:
> > There is no cloud user on the host/agent side. Agent runs as root. Or is
> > this an experiment to try to run the agent as another user?
>
> That was entirely my mistake. I just went about thinking the agents
> ran as cloud (ssvm, cpvm, kvm agent). Looking at my deployment it
> seems that /etc/sudoers has an entry for cloud but the service runs as
> root. My apologies for misleading on that.
>
> > On Jul 23, 2013 4:44 AM, "Prasanna Santhanam" <t...@apache.org> wrote:
> >
> > > On Tue, Jul 23, 2013 at 06:19:06PM +0800, Indra Pramana wrote:
> > > > Dear Prasanna and all,
> > > >
> > > > I have tried to remove the host from CloudStack, uninstall and
> reinstall
> > > > cloudstack-agent and added the host back into CloudStack. The "cloud"
> > > > user-id is still not yet created. I then tried to add the "cloud"
> user
> > > > manually, using exactly the same credentials as the "cloud" user on
> the
> > > > other host.
> > > >
> > > > >From the host, I tried to do a virsh connect qemu+ssh to the other
> host
> > > > using the "cloud" user (instead of root), and getting this error:
> > > >
> > > > cloud@hv-kvm-01:~$ virsh --connect qemu+ssh://
> cloud@10.237.3.22/systemlist
> > > > cloud@10.237.3.22's password:
> > > > error: failed to connect to the hypervisor
> > > > error: no valid connection
> > > > error: End of file while reading data: : Input/output error
>
> However - the issue still remains that migration between two KVM hosts
> seems to be failing. The log indicates the error shown above which was
> tried by hand here.
>
> Deployment contains KVM hosts with Ceph+RBD as primary storage.
>
> --
> Prasanna.,
>
> ------------------------
> Powered by BigRock.com
>
>
