Date: Wed, 14 Mar 2012 08:34:26 -0400 From: Mark Murawski <markm-lists@intellasoft.net> To: freebsd-fs@freebsd.org Subject: Re: ZFS file corruption problem Message-ID: <4F609052.5010300@intellasoft.net> In-Reply-To: <20120314105011.Horde.mYG5YpjmRSRPYGnT6OFEHuA@webmail.leidinger.net> References: <4F5F7116.3020400@intellasoft.net> <4F5F97A4.6070000@brockmann-consult.de> <4F60266D.1090302@intellasoft.net> <4F6027B3.5080006@intellasoft.net> <20120314105011.Horde.mYG5YpjmRSRPYGnT6OFEHuA@webmail.leidinger.net>
next in thread | previous in thread | raw e-mail | index | archive | help
On 03/14/2012 05:50 AM, Alexander Leidinger wrote: > Quoting Mark Murawski <markm-lists@intellasoft.net> (from Wed, 14 Mar > 2012 01:08:03 -0400): > >> On 03/14/12 01:02, Mark Murawski wrote: > >>> Why would the whole pool now become available upon access to a bad file? > > Because you configured it like this (respectively didn't configure a > different behavior). > >> Also... isn't this pretty terrible behavior that the process accessing >> the bad file is unkillable? > > If you are in an environment where the disks are not local (ZFS is > designed with corporate environments in mind), you do not want to fail > on an application level or panic because of a small hickup in the network. > > man zpool: > ---snip--- > failmode=wait | continue | panic > Controls the system behavior in the event of catastrophic pool fail? > ure. This condition is typically a result of a loss of connectivity > to the underlying storage device(s) or a failure of all devices > within the pool. The behavior of such an event is determined as fol? > lows: > > wait Blocks all I/O access until the device connectivity is recov? > ered and the errors are cleared. This is the default behav? > ior. > > continue > Returns EIO to any new write I/O requests but allows reads to > any of the remaining healthy devices. Any write requests that > have yet to be committed to disk would be blocked. > > panic Prints out a message to the console and generates a system > crash dump. > ---snip--- > > It is up to you to switch to 'continue' or 'panic' for local disks. > > Bye, > Alexander. > Oh... wow. It's not that I've configured if that way in particular, it's more of a matter of the default settings came like that. But anyway, thanks a ton. I had no idea that was configurable. I was even thinking that "you know, it would be nice if that behavior was configurable". Once I started running into these problems I started losing faith in zfs and its design. I've been dealing with this corrupt files problem for about a week now. Finding out the fix was as simple as deleting the file and setting a new config option has reaffirmed my belief in zfs.
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?4F609052.5010300>