Thanks to both John and Brian.

On 12/2/06 1:34 AM, "John R Pierce" pierce@hogranch.com wrote:

find a post from me earlier today, I give all the details of what worked with centos 4.x

I hadn't gotten too reading any of the lists inbox at that point. Yours was a good email with good detail, my timing was obviously off <grin>.

I've done some reasonable testing and found it to be nearly the speed of local disk on my test system. I have a rolled ethernet cable between the initiator and the target and they happily negotiate 1000bTX-FD.

I manually mounted the ext3fs and ran my tests really without issue. If the target goes away, the initiator sits and waits like it would for a physical drive -- as I would expect. I then simulated a full powerloss situation and ended up where the initiator was up and running prior to the target. The iscsi initiator waited for a while eventually to continue with a soft failure, but my initiator stopped booting as the entry in /etc/fstab saw this as a broken drive.

As none of the functional OS is on the iscsi drive, is there an option to include in /etc/fstab to make this a soft error and let the system resume booting? In the world of NFS I'd use the 'intr' option but I'm not sure what to do beyond using 'noauto' and then have a late script try to mount it manually just before the apps that use that partition.

Is there a common/popular way to workaround this situation? I'd guess that FCAL devices have to cope with this too -- I have no experience there though.

Thanks,

peter