Skip site navigation (1)Skip section navigation (2)
Date:      Sat, 12 Jun 1999 08:02:31 -0700
From:      Greg Lehey <grog@lemis.com>
To:        Joe McDonald <joe@vpop.net>, Chris Cook <ccook@tcworks.net>
Cc:        freebsd-isp@freebsd.org
Subject:   Re: redunancy without RAID
Message-ID:  <19990612080231.64973@folly.lemis.com>
In-Reply-To: <37683170.81567132@127.0.0.1>; from Joe McDonald on Fri, Jun 11, 1999 at 08:57:40AM -0700
References:  <Pine.GSO.3.96.990602155911.20098A-100000@jane.lfn.org> <37562454.718D3319@tcworks.net> <37683170.81567132@127.0.0.1>

next in thread | previous in thread | raw e-mail | index | archive | help
On Friday, 11 June 1999 at  8:57:40 -0700, Joe McDonald wrote:
> On Thu, 03 Jun 1999 01:44:36 -0500, Chris Cook <ccook@tcworks.net> wrote:
>
>>        We have been running VINUM here at our ISP for quite some time
>> now.  It works great... we have our /usr and /var partitions setup on 2
>> different mirrored volumes.  The overhead is low and access times are
>> great.  With VINUM, if one of the mirrored disks go out, the volume should
>> not crash <grin>.  As far as ccd goes, I have never used it, but I believe
>> both ccd and VINUM will not mirror a boot partition (I know VINUM won't
>> yet).  The VINUM was difficult to install, only because I was trying to use
>> the port that comes with FBSD 3.1.

I don't understand this.  Vinum is part of the base system in 3.X.

>> If you go this route, I suggest downloading the lastest version and
>> you shouldn't have problems.  Hope this helps...

It's certainly a good idea to get the latest version.  A lot of
changes went in until about mid-April.

> Do you know the best way to test if the mirror really works if a drive goes
> bad?  Pull the power to the drive as it's running?  Pull the scsi cable to the
> drive as it's running?  Will this cause physical damage to the drive?

I try both pulling the power cable and pulling the SCSI cable.  It's
possible that you could confuse the system if you pull the data cable
during a transfer, but it hasn't happened to me yet, and anyway, this
is a test environment.

The way Vinum handles these errors, there are really only two errors
that count:

1.  Drive goes away.  CAM reports "invalidating pack", and Vinum takes
    the drive down and all associated subdisks into the "crashed"
    state.  If the drive comes back before any writes are attempted,
    the drive goes into a "reborn" state if there are other plexes, or
    an "up" state if there are none.  In "reborn", Vinum will write to
    the subdisk, but it won't read it if it can get the data
    elsewhere.

2.  I/O error on a transfer.  Vinum marks the subdisk "stale".

In either case, what happens to the plexes is relatively complex.
Normally they're marked "degraded".  The volumes stay up as long as
*anything* is accessible (since otherwise you'd get I/O errors when
you tried to access them).

Greg
--
Finger grog@lemis.com for PGP public key
See complete headers for address and phone numbers


To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-isp" in the body of the message




Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?19990612080231.64973>