🚀 go-pugleaf

RetroBBS NetNews Server

Inspired by RockSolid Light RIP Retro Guy

Article View: comp.os.netware.misc
Article #451

Re: Doing VREPAIRS and restores while re-mirroring

#451
From: David Gersic
Date: Tue, 02 Apr 1996 00:00
84 lines
4828 bytes
In <g10XxA1M1yWC088yn@winternet.com>, james@winternet.com (James Borchart) writes:
>In article <4j9elv$pc0@corn.cso.niu.edu>,
>dgersic@niu.edu (David Gersic) wrote:
>>
>>[...] During one of the
>>BrainShare sessions I was at last year, the presenter explicitly
>>recommended removing the mirror, then running vrepair, and re-mirroring the
>>disks once everything was ok.
>>
>>
>>When running VREPAIR, recommend breaking the mirror on mirrored/duplexed
>>  volumes. Fix one set, then re-mirror. Can use the broken mirror to try
>>  other recovery options if needed. Also halves the disk i/o required, so
>>  should run faster. Remirror after repair.
>>
>This is nothing short of insanity, and it is unfortunate that it was posted.

I don't agree, or disagree, with the presenter, but I do tend to believe
that he knows what he's talking about.

>One of the fundamental purposes of mirroring is TO HAVE ENOUGH INFORMATION
>TO RECREATE ONE LOGICAL DRIVE FROM TWO PHYSICAL DRIVES.  Never, ever break
>the mirror except as a last resort.

To elaborate on the original post, the presenter's point was that if you
break the mirror before attempting other repairs that you *have* another
option if the vrepair fails; you can ship the drive to On Track or similar
for data recovery. Once you start trying to use the logical repair
(vrepair) process, if it doesn't work, you have no other options as the bad
"corrected" information will have been written to both drives.

>Lets say that Fat sector one is busted
>on drive 0's first volume, and fat sector two is busted on drive 1's first
>volume, and that 0 and 1 are mirrored together.  If the mirror were broken and
>the drives vrepair'ed separately, then it is possible that both volumes
>would get the dreaded "hole in fat" error.  However, if they were
>vrepaired together, then you would have a complete drive when done.

Maybe, if you're lucky. OTOH, if you have a disk driver problem (not a
physical error, a software bug) that trashes one FAT, but doesn't manage to
trash the mirror's copy of the FAT, if vrepair can't fix it, it will also
screw up the mirror's copy. Neither scenario is a happy one, and I hope to
avoid having either one happen to me.

Mirroring has saved my a** from hardware errors. It cannot save me from
logical/software errors.

>There is no logic to breaking the mirror for a vrepair.  There is no point
>in having mirrored drives if you break the mirror for vrepair.  Only break
>the mirror in the event of a partial or complete hardware failure on the drive.

I think that this would depend, very heavily, on *exactly* what problem was
being worked on. For a physical failure, there isn't much point in keeping
a dead drive in the mirrored set. For a partial physical failure, hot-fix
should handle all (maybe only most) of the problem. Software failures are
the ones the presenter was talking about, and those are harder to diagnose
and repair. He was also giving some backround on how vrepair works,
including the algorthm it uses to decide which parts of conflicting
information is "correct" as it repairs the volume.

>Remember that once a mirror is broken, it CAN NOT be put back together, you
>CAN NOT go back.

Right. Maybe I should have elaborated more in the first post. He was not
talking about anything that the average (or even above average) person
would be able to accomplish.

>I am not saying that David is confused or misinterpreting what he heard,
>I am merely saying that the information he posted is the exact opposite
>of correct.

Right, I'm neither confused or misinterpretting anything; I understand
exactly what he was talking about. I may have been unclear in my post, and
you may have had to be there in the session for it to make sense.


======================================================================

    The packet goes out the card, into the copper, out the router,
    onto the fiber, across the world, thru the copper............
    NOTHING BUT NET.

David Gersic                                           dgersic@niu.edu
Systems Programmer                        Northern Illinois University
Technical Details
Message-ID: <4jsai4$lof@corn.cso.niu.edu>
Path: arc1-ca.pugleaf.net!unknown!not-for-mail
References: <31571b42.959100@news.execpc.com> <4j9elv$pc0@corn.cso.niu.edu> <g10XxA1M1yWC088yn@winternet.com>

Show Full Headers
Complete Article Headers
X-Google-Language: ENGLISH,ASCII-7-bit
X-Google-Thread: fe659,3984ce55a85bd1f6
X-Google-Attributes: gidfe659,public
X-Google-Thread: f8bf7,3984ce55a85bd1f6
X-Google-Attributes: gidf8bf7,public
From: dgersic@niu.edu (David Gersic)
Subject: Re: Doing VREPAIRS and restores while re-mirroring
Date: 1996/04/02
Message-ID: <4jsai4$lof@corn.cso.niu.edu>
X-Deja-AN: 145495371
references: <31571b42.959100@news.execpc.com> <4j9elv$pc0@corn.cso.niu.edu> <g10XxA1M1yWC088yn@winternet.com>
organization: Northern Illinois University
reply-to: dgersic@niu.edu (David Gersic)
newsgroups: comp.os.netware.misc,bit.listserv.novell