[ovirt-users] iSCSI Multipathing -> host inactive

[Uwe Laverenz]

Hi,

Am 16.08.2016 um 09:26 schrieb Elad Ben Aharon:
> Currently, your host is connected through a single initiator, the
> 'Default' interface (Iface Name: default), to 2 targets: tgta and tgtb

I see what you mean, but the "Iface Name" is somewhat irritating here, 
it does not mean that the wrong interface (ovirtmgmt) is used.
If you have a look at "Iface IPaddress" for both you can see that the 
correct, dedicated interfaces are used:

Iface IPaddress: 10.0.131.122   (iSCSIA network)
Iface IPaddress: 10.0.132.122   (iSCSIB network)

> (Target: iqn.2005-10.org.freenas.ctl:tgta and Target:
> iqn.2005-10.org.freenas.ctl:tgtb). Hence, each LUN is exposed from the
> storage server via 2 paths.
> Since the connection to the storage is done via the 'Default' interface
> and not via the 2 iSCSI networks you've configured, currently, the iSCSI
> bond is not operational.

Please see above. The storage servers iSCSI-addresses aren't even 
reachable from the ovirtmgmt net, they are in completely isolated networks.

> For the iSCSI bond to be operational, you'll have to do the following:
> - Create 2 networks in RHEVM under the relevant cluster (not sure if
> you've already did it) - iSCSI1 and iSCSI2 . Configure both networks to
> be non-required networks for the cluster (should be also non-VM networks).
> - Attach the networks to the host's 2 interfaces using hosts Setup-networks.
> - Create a new iSCSI bond / modify the bond you've created and pick the
> 2 newly created networks along with all storage targets. Make sure that
> the Default network is not part of the bond (usually, the Default
> network is the management one - 'ovirtmgmt').
> - Put the host in maintenance and re-activate it so the iSCSI sessions
> will be refreshed with the new connection specifications.

This is exactly what I did, expect that I had to add the iSCSI-storage 
first, otherwise the "iSCSI Multipathing" tab does not appear in the 
data center section.

I configured an iSCSI-Bond and the problem seems to be that it leads to 
conflicting iSCSI-settings on the host. The host uses the very same 
interface twice only with different "IFace Name":

iSCSIA:

Iface Name: default
Iface Transport: tcp
Iface Initiatorname: iqn.1994-05.com.redhat:cda91b279ac5
Iface IPaddress: 10.0.131.122

Iface Name: enp9s0f0
Iface Transport: tcp
Iface Initiatorname: iqn.1994-05.com.redhat:cda91b279ac5
Iface IPaddress: 10.0.131.122


iSCSIB:

Iface Name: default
Iface Transport: tcp
Iface Initiatorname: iqn.1994-05.com.redhat:cda91b279ac5
Iface IPaddress: 10.0.132.122

Iface Name: enp9s0f1
Iface Transport: tcp
Iface Initiatorname: iqn.1994-05.com.redhat:cda91b279ac5
Iface IPaddress: 10.0.132.122

I guess this is the reason why the host has problems to attach the 
storage domain, it toggles all storage domains on and off all the time.

Thank you,
Uwe