Thanks, Mike for the valuable information. I will try both: dm-multipath
and the redirection.


On 05/08/11 01:03, Mike Christie wrote:
> On 08/02/2011 06:19 AM, Sebastian Riemer wrote:
>   
>> This is the sequence:
>> 1. pause the VM
>> 2. unlink replication to new location, new target becomes primary
>> 3. logout from old iSCSI target
>> 4. login to new iSCSI target, BUT with the same block device as it is
>> held open
>>     
> The problem with 3 and 4 and using the iscsi disk directly and not using
> something like iscsi login redirect, is that on logout the device
> (/dev/sdXYZ) is going to be destroyed, so if something was referencing
> it they are basically screwed. All further access with result in IO errors.
>
> When you do #4 and login you get a new /dev/sdXYZ and so your app has to
> open that and use it.
>
> If your setup can just use dm-multipath then you can just multipath over
> iscsi. When you logout it would just destroy what the multipath device
> sees as a path. Then when you login, it will add it as a new path. You
> can then resume the VM and it would never know that the paths under the
> multipath device got swithced around.
>
>   
>> Does iSCSI target redirection work without stopping VMs?
>>     
> Yeah, it should. But we can only redirect to another portal. If you are
> looking at the open iscsi userspace source code see
> usr/login.c:iscsi_update_address and initiator.c's
> __check_iscsi_status_class's ISCSI_LOGIN_STATUS_TGT_MOVED_PERM handling.
>
> So the target name would have to be the same but the portal info could
> be different.
>   

-- 
You received this message because you are subscribed to the Google Groups 
"open-iscsi" group.
To post to this group, send email to open-iscsi@googlegroups.com.
To unsubscribe from this group, send email to 
open-iscsi+unsubscr...@googlegroups.com.
For more options, visit this group at 
http://groups.google.com/group/open-iscsi?hl=en.

Reply via email to