Adding new iSCSI Volumes

[the_nipper]

This is exactly what I have with my DSS here:

There are around 15 iSCSI targets running. I just create a new iSCSI _volume_ (not yet assigning a target!), I get the warnings about connection reset, and then the current iSCSI initiators get lots problems like

Feb 10 14:01:47 vms-4035 vmkernel: 9:02:24:49.578 cpu4:1104)iSCSI: session 0xb71
2370 to iqn.shoe-2013.vmachines-sb dropped

This was a hangup of VMware ESX iSCSI target, where all vmachines reside. I have to reboot. I found this problem several times.

It is absolutely unacceptable that the creation of a target volume causes running iSCSI connections to fail. Storage software designed for TB of space must be able to handle this without timeouts etc! I can't stop my company every time I need some 100 GB of space.

Regards,

Robert

[budy]

Hmm, I do think that your initiators are to blame then. I am using also multiple targets and my initiators don't seem to have troubles with that.

In fact I would expect an iSCSI initiator to be able to cope with short "outages" and to reconnect to the targets on its own.

Cheers,
budy

[zeki893]

Yes I have the same problem as the_nipper. It is unacceptable to have to shut everything down. When my vm reconnects to the drive it becomes read only mode and requires a restart to become writable again. I have moved to nfs or nas because of this.

[goranl]

Does anyone have a definitive awnser to this one?

I'm dealing with the same issue here. I can't imagine I have to close down my 19 VL's in order to create a new one.

I put in a question via the support desk. I'll keep you posted.

[Robotbeat]

Yeah, this is a huge problem. It has to be fixed. We have to know a firm ETA about this. Put this at the top of your queue, because this puts a huge limit on the type of environments this can be deployed under, but I'm sure you already know that. We all need a fix for this as soon as possible.

We're having this problem with fibre channel.

[Robotbeat]

There has to be a way of fixing this because EMC and folk don't have this problem.

[Robotbeat]

BTW, this is not an issue if you publish all of your logical volumes to the default group in fibre channel. In that case, you would have to do LUN masking on the INITIATOR side instead of on the Open-E TARGET side. If you have QLogic initiator HBAs in Windows, then use the QLconfig utility to select the luns that should be used on each initiator. This is a pain because you have to do this on every initiator machine, but it should work.

Also, you can use the QLDirect driver for fibre channel path failover (like MPIO but no active-active, it's active-passive). This is I think the only way to do fibre channel (path) failover in Windows 2003 or earlier. (Windows 2008 has a generic Fibre channel MPIO driver, I think.)