Skip site navigation (1)Skip section navigation (2)
Date:      Sat, 30 Apr 2011 14:42:41 -0300
From:      Chris Forgeron <cforgeron@acsi.ca>
To:        Jeremy Chadwick <freebsd@jdc.parodius.com>, Volodymyr Kostyrko <c.kworr@gmail.com>
Cc:        "freebsd-fs@freebsd.org" <freebsd-fs@freebsd.org>
Subject:   RE: ZFS v28 for 8.2-STABLE
Message-ID:  <BEBC15BA440AB24484C067A3A9D38D7E0163C7730518@server7.acsi.ca>
In-Reply-To: <20110430001524.GA58845@icarus.home.lan>
References:  <4DB8EF02.8060406@bk.ru> <ipf6i6$54v$1@dough.gmane.org> <20110430001524.GA58845@icarus.home.lan>

next in thread | previous in thread | raw e-mail | index | archive | help
IMHO:

I've been working with ZFS v28 since the early days that pjd made the patch available, and while I'm actually using it daily in (a very limited and controlled) production, I feel it needs more testing before it's under serious consideration to be included in Stable. 

It's got a few little catches and gotchas that will frustrate anyone who hasn't been working with it - Hangs here and there, and the potential to destroy your pool if you're not careful. Just last night, I added a spare drive to a 10 drive pool (two five-drive raidz striped) and it hung on me. After 20 minutes of waiting, I hard rebooted, and issued the command again, and it completed within 30 seconds.  Was I suffering resource exhaustion someplace? It's hard to say, as it happened once and it's tricky to reproduce.  This box was running for over 30 days, serving a lot of disk-io to an ESX cluster. 

This isn't a slight against pjd's work at all - He's done an excellent job, and the issues I'm running into aren't easily reproducible by others. It's up to me to take the time to trace the issues down so they can be properly squashed, but I'm busy putting out other fires for now.

We all want to see v28 with the stamp of approval that Stable will give it, but I think that event is well into the summer months. 

Let's get more people stress testing it and trying it on different configurations, hopefully that will give the feedback that pjd needs for final tweaking and correction. 

If anything, make the v28 patchfile very easy to obtain and install - or possibly a kernel option like the old/new NFS server, so people can enable the new v28 code easier for testing purposes?




Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?BEBC15BA440AB24484C067A3A9D38D7E0163C7730518>