Skip site navigation (1)Skip section navigation (2)
Date:      Thu, 06 Jul 2006 09:34:40 -0400
From:      Chuck Swiger <cswiger@mac.com>
To:        Jeremy Ehrhardt <jeremye@caltech.edu>
Cc:        freebsd-questions@freebsd.org
Subject:   Re: vinum stability?
Message-ID:  <44AD1170.1000205@mac.com>
In-Reply-To: <44AC5190.9070001@caltech.edu>
References:  <44AC5190.9070001@caltech.edu>

next in thread | previous in thread | raw e-mail | index | archive | help
Jeremy Ehrhardt wrote:
> We've been testing this box as a 
> file server, and it usually works fine, but smartd reported a few bad 
> sectors on one of the drives, then a few days later it crashed while I 
> was running chmod -R on a directory on "drugs" and had to be manually 
> rebooted. I can't figure out exactly what happened, especially given 
> that RAID 5 is supposed to be robust against single drive failures and 
> that despite the bad blocks smartctl claims the drive is healthy.

As soon as you notice bad sectors appearing on a modern drive, it's time to 
replace it.  This is because modern drives already use spare sectors to 
replace failing data areas transparently, and when that no longer can be done 
because all of the spares have been used, the drive is likely to die shortly 
thereafter.

RAID-5 provides protection against a single-drive failure, but once errors are 
seen, the RAID-volume is operating in degraded mode which involves a 
significant performance penalty and you no longer have any protection against 
data loss-- if you have a problem with another disk in the meantime before the 
failing drive gets replaced, you're probably going to lose the entire RAID 
volume and all data on it.

> I have three questions:
> 1: what's up with gvinum RAID 5? Does it crash randomly? Is it 
> considered stable? Will it lose data?

Gvinum isn't supposed to crash randomly, and it reasonably stable, but it 
doesn't seen to be as reliable as either a hardware RAID setup or the older 
vinum from FreeBSD-4 and earlier.

As for losing data, see above.

> 2: am I using a SATA controller that has serious problems or something 
> like that? In other words, is this actually gvinum's fault?

If you had a failing drive, that's not gvinum's fault.  gvinum is supposed to 
handle a single-drive failure, but it's not clear what actually went 
wrong...log messages or dmesg output might be useful.

> 3: would I be better off using a different RAID 5 system on another OS?

Changing OSes won't make much difference; using hardware to implement the RAID 
might be an improvement, rather than using gvinum's software RAID.  Of course, 
you'd have to adjust your config to fit within your hardware controller's 
capabilities.

-- 
-Chuck



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?44AD1170.1000205>