[ovirt-users] ovirt don't give "initiator name" to cinder when I'm launching a VM with a cinder volume

Nir Soffer nsoffer at redhat.com
Wed Mar 8 11:18:09 UTC 2017


On Wed, Mar 8, 2017 at 10:04 AM, herve <rv.morin.93 at gmail.com> wrote:
> dear ovirt user community,
>
> I can't sort out an ovirt issue I got with cinder for weeks, I hope someone
> could give me a clue...
> From mi ovirt manager 4.1, I can create and delete a cinder volume or attach
> and detach a cider vomume from an ovirt  VM.
>
> The problem is I never had been able to launch an ovirt VM with a cinder
> volume attached to it.
> I just got a simple message from ovir web gui:  "Failed to run VM
> centoscinder (User: admin at internal-authz)".
>
> what i am using is
> a RDO newton release of openstack. the backend of cinder is a local lvm,

Here is your problem, ovirt supports only ceph storage with cinder.

We may add more storage types in the future.

Nir

> and
> the openstack VMs can use the cinder volumes from other compute node.
> from the ovirt 4.1 fedora 24 server:
> [root at rhevf24 ~]# rpm -qa | grep ovirt-engine-4
> ovirt-engine-4.1.0.4-1.fc24.noarch
> ovirt-engine-setup-plugin-ovirt-engine-4.1.0.4-1.fc24.noarch
> [root at rhevf24 ~]#
>
> from the ovirt server , through the iscsiadm command,  I can see for example
> the 2 disks attached to 2 diffents  openstack instances,
> but not the one I just created from ovirt which is attached to nothing ,
> from the openstack rdo point of view.
> [root at rhevf24 iscsi]# iscsiadm -m discovery -t sendtargets -p 192.168.122.36
> 192.168.122.36:3260,1
> iqn.2010-10.org.openstack:volume-f4c6c55c-e383-4781-8d24-d41b55870d29
> 192.168.122.36:3260,1
> iqn.2010-10.org.openstack:volume-58844340-741e-45da-9aa7-b0ed08391df3
>
> as I said previously From ovirt I can create /delete in cinder and
> attach/detach volumes to ovirt VMs I don't get any troubles with that.
>
> ----
> in ovirt engine.log almost nothing , except it can't aquire lock.. but in
> openstack cinder, and with tcpdump I get more informations.
> tcpdump give me when ovirt intent the iscsi connection something like:
> json.key == "os-initialize_connection"
> {.  "os-initialize_connection" : {.    "connector"  :  {  }, },}
>
> and I get from "tail -f  /var/log/cinder/volume.log" :
> 2017-03-07 16:45:06.889 3856 ERROR cinder.volume.targets.iscsi
> [req-b6d39a14-f9ea-4552-91e7-5db811217b1e dd1dd33789274fb7ba228c75a07c0e94
> 0f68893965dd4f5dafd319cbffb33bb4 - default default] The volume driver
> requires the iSCSI initiator name in the connector.
> ...
> ...
> 2017-03-07 16:45:06.891 3856 ERROR oslo_messaging.rpc.server InvalidInput:
> Invalid input received: Connector doesn't have required information:
> initiator
>
>
> So it looks like I didn't get  or I can't provide the initiator name I want
> to be connected to.
> Cinder can't obviously guess which one I want. that's more or less what's
> the cinder's logs are telling me.
>
> I have searched for a long time in internet, it looks like  at least since
> ovirt 3.6, cinder is ont anymore an issue, and can even be integrated with a
> ceph back end to ovirt...
> cinder model an client are installed in the ovirt server...
>
> Somebody get any clue to tell ma where i made a mistake in the cinder
> integration with ovirt , which seems to be correct regarding the
> create/delete attach /detach volume behavior...
>
> Thank you
>
> Herve
>
>
>
>
>
>
>
>
> _______________________________________________
> Users mailing list
> Users at ovirt.org
> http://lists.ovirt.org/mailman/listinfo/users


More information about the Users mailing list