Date: Sun, 24 Mar 2013 08:54:48 -0700 From: Jeremy Chadwick <jdc@koitsu.org> To: Quartz <quartz@sneakertech.com> Cc: freebsd-fs@freebsd.org Subject: Re: ZFS question Message-ID: <20130324155448.GA4122@icarus.home.lan> In-Reply-To: <20130324153342.GA3687@icarus.home.lan> References: <20130321044557.GA15977@icarus.home.lan> <514AA192.2090006@sneakertech.com> <20130321085304.GB16997@icarus.home.lan> <20130324153342.GA3687@icarus.home.lan>
next in thread | previous in thread | raw e-mail | index | archive | help
On Sun, Mar 24, 2013 at 08:33:42AM -0700, Jeremy Chadwick wrote:
> However, commands like "zpool status"
...and seems a typo I made in vim caused the rest of my sentence to get
deleted before I sent it out. This should have read:
> However, commands like "zpool status" work just fine, but things like
> "zpool destroy" and so on indefinitely block ("mount drain"), which to
> me makes some degree of sense.
To expand: for example, you've lost 3 disks of a 4-disk raidz2 pool,
your data is buggered and you'll need to recover from backups.
Yes, you will need to reboot for the ZFS layer to effectively "un-wedge"
itself from whatever catatonic state its in. No argument: this is a bug
somewhere, and my guess is that it relates to the confused state of the
devices in CAM-land. But regardless, I think if you were to lose 3 of 4
disks on a raidz2 pool you'd have much more serious things to be worried
about than "well crap I have to issue a reboot".
And yes, I did test a reboot in the scenario I described -- the system
did reboot without physically pressing the button. But then again, for
remotely-managed systems, administrators should have the ability to
remotely power-cycle or force resets (e.g. drop to DDB via serial
console and force a reset).
People who run servers remotely yet lack this capability are
intentionally choosing to live dangerously and I do not condone such.
These folks also make me wonder how they update world without remote
console access, since to do it right you *must* drop to single-user for
the installworld phase.
I learned my lesson of "assuming" installworld would work from
multi-user long ago when it broke one time and I ended up with a system
with broken /libexec/ld-elf* binaries. Having to go to the datacenter
30 minutes away at 3 in the morning taught me to follow instructions.
--
| Jeremy Chadwick jdc@koitsu.org |
| UNIX Systems Administrator http://jdc.koitsu.org/ |
| Mountain View, CA, US |
| Making life hard for others since 1977. PGP 4BD6C0CB |
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20130324155448.GA4122>