>Well, I can tell you what I think: > >WOOOOOOHHHHHHOOOOOOOOOOOOOOOOOOO!!! > >I haven't had time to read that entire discussion yet (gotta take >the baby & the dog out for a walk), but I can't tell you guys how >happy I am to see this. I'm desperate to use ZFS over iSCSI to >mirror a couple of servers, but it was an unworkable solution with a >3 minute timeout. Being able to tweak that down to a 1-2 second >timeout would be great. Funnily enough, exactly the opposite >situation to yours; I want it far less tolerant than the default. > >Could I ask you to please touch base with the iSER guys (Peter >Dunlap, David Hollister & Jesse Butler) since I was e-mailing them >only last week to find out whether this had been addressed in their >code. At the time it hadn't since it's based on the standard iSCSI >code, but if that means these parameters can be integrated there too >it would be perfect. > >For me, what needs tweaking is whatever parameters define the >timeout period when a cable is disconnected, since ZFS hangs the >whole pool for 3 minutes waiting for a reply if an iSCSI device goes >offline. I believe that's line 227 of iscsi.h, but I'm not 100% >sure: >http://bugs.opensolaris.org/view_bug.do?bug_id=6670866 >
I'm looking to implement a very similar setup, and I wholeheartedly endorse reducing this particular timeout to a few seconds! A very simple solution, by the way, might be to provide a property such as the failmode parameter for zpools with its option of "continue, wait, panic" where "continue" utilizes a very short timeout. -- Maurice Volaski, [EMAIL PROTECTED] Computing Support, Rose F. Kennedy Center Albert Einstein College of Medicine of Yeshiva University _______________________________________________ storage-discuss mailing list [email protected] http://mail.opensolaris.org/mailman/listinfo/storage-discuss
