Dying Disks on Mirrored Win 2003

Have a 2003 Dell Server with 8 bays

The O/S C:\ drive and data drive E:\ are mirrored.
Everything is on 145G Seagate Cheetah (or Barracuda) SCSI disks.

Somehow occasionally, the mirror (secondary ones) do orphan - never the
Original (primary disks) on both C:\ and E:\
50% of the times, I reactivate the disk, resynch and done.

But in half of the cases, I cannot reactivate the disk, it is totally dead.
I cannot reformat and reuse.

The issue is why it is "always" the mirrored secondary disk.
And also, why they die totally without any physical harm, to the extend that
they have to be destroyed or sent to Seagate.

Is this is common thing with Mirroring? I haven't seen this in my previous
life with 3-4 years on different servers.

Travis McGee <> wrote:
> Have a 2003 Dell Server with 8 bays
>
> The O/S C:\ drive and data drive E:\ are mirrored.
> Everything is on 145G Seagate Cheetah (or Barracuda) SCSI disks.
>
> Somehow occasionally, the mirror (secondary ones) do orphan - never
> the Original (primary disks) on both C:\ and E:\
> 50% of the times, I reactivate the disk, resynch and done.
>
> But in half of the cases, I cannot reactivate the disk, it is totally
> dead. I cannot reformat and reuse.
>
> The issue is why it is "always" the mirrored secondary disk.
> And also, why they die totally without any physical harm, to the
> extend that they have to be destroyed or sent to Seagate.
>
> Is this is common thing with Mirroring? I haven't seen this in my
> previous life with 3-4 years on different servers.

Not sure, but I don't use or recommend software mirroring. I use hardware
RAID only on production servers - it is much more reliable, has more
features/options, and performs a lot better.

I am very well aware of the pros and cons of hardware vs. software
mirroring.
Everytime I open my mouth about mirroring, people always pitch Hardware
Mirroring.
The problem is that - I do software mirroring for reasons that are specific
to this server.

So the question is still outstanding..... and ideas?

"Bill Kearney" <> wrote in message
news: t...
>> Not sure, but I don't use or recommend software mirroring. I use hardware
>> RAID only on production servers - it is much more reliable, has more
>> features/options, and performs a lot better.
>
> Hardware RAID has it's own share of problems. Namely you can't usually
> take a drive off one and use it in another machine. Even if you've got
> the same RAID hardware in the other box. Even slight variations in
> firmware often spell disaster for trying to recover from a hardware
> failure on the host motherboard. At least with software RAID you can
> bring the drive(s) up on another box, using entirely different hardware
> and often a different OS rev, and get to the data.
>
> Both have their issues, but it's wrong to say one is 'better' than the
> other.
>
> -Bill Kearney
>

I don't think the RAID or how it is done has anything to do with the failure
rate of the drives. The drives are failing just simply because the drives
are failing. You could be using them in a machine with no RAID at all
whatsoever and would experience the same thing.

Maybe you just been unlucky with the drive purchases.
Maybe your machine is in an area where the temperature is too high.

--
Phillip Windell

The views expressed, are my own and not those of my employer, or Microsoft,
or anyone else associated with me, including my cats.
-----------------------------------------------------


"Travis McGee" <> wrote in message
news:%...
>I am very well aware of the pros and cons of hardware vs. software
>mirroring.
> Everytime I open my mouth about mirroring, people always pitch Hardware
> Mirroring.
> The problem is that - I do software mirroring for reasons that are
> specific to this server.
>
> So the question is still outstanding..... and ideas?
>
> "Bill Kearney" <> wrote in message
> news: t...
>>> Not sure, but I don't use or recommend software mirroring. I use
>>> hardware RAID only on production servers - it is much more reliable, has
>>> more features/options, and performs a lot better.
>>
>> Hardware RAID has it's own share of problems. Namely you can't usually
>> take a drive off one and use it in another machine. Even if you've got
>> the same RAID hardware in the other box. Even slight variations in
>> firmware often spell disaster for trying to recover from a hardware
>> failure on the host motherboard. At least with software RAID you can
>> bring the drive(s) up on another box, using entirely different hardware
>> and often a different OS rev, and get to the data.
>>
>> Both have their issues, but it's wrong to say one is 'better' than the
>> other.
>>
>> -Bill Kearney
>>
>
>

Bill Kearney <> wrote:
>> Not sure, but I don't use or recommend software mirroring. I use
>> hardware RAID only on production servers - it is much more reliable,
>> has more features/options, and performs a lot better.
>
> Hardware RAID has it's own share of problems. Namely you can't
> usually take a drive off one and use it in another machine. Even if
> you've got the same RAID hardware in the other box. Even slight
> variations in firmware often spell disaster for trying to recover
> from a hardware failure on the host motherboard. At least with
> software RAID you can bring the drive(s) up on another box, using
> entirely different hardware and often a different OS rev, and get to
> the data.
> Both have their issues, but it's wrong to say one is 'better' than the
> other.
>
> -Bill Kearney

<flamewar mode>

You have valid points, but I still say hardware RAID is the way to go. I
don't need to take a drive off one server and mount it in another. I've got
good backups, I've got as much redundant hardware in the server as I can
throw at it, and I use global hotspares so I can handle two drive failures
without the machine puking up blood.

</flamewar mode>

;-)

Travis McGee <> wrote:
> I am very well aware of the pros and cons of hardware vs. software
> mirroring.
> Everytime I open my mouth about mirroring, people always pitch
> Hardware Mirroring.

Yes. Caveat emptor in usenet.

> The problem is that - I do software mirroring for reasons that are
> specific to this server.

Fair enough. Mentioning that right off the bat might preclude some people
(myself included) from barging in with unwanted advice. Most people who come
into the groups with software RAID problems are less informed than you may
be.
>
> So the question is still outstanding..... and ideas?

It sounds like you've got bum hardware, honestly, as Phil pointed out.

>
> "Bill Kearney" <> wrote in message
> news: t...
>>> Not sure, but I don't use or recommend software mirroring. I use
>>> hardware RAID only on production servers - it is much more
>>> reliable, has more features/options, and performs a lot better.
>>
>> Hardware RAID has it's own share of problems. Namely you can't
>> usually take a drive off one and use it in another machine. Even if
>> you've got the same RAID hardware in the other box. Even slight
>> variations in firmware often spell disaster for trying to recover
>> from a hardware failure on the host motherboard. At least with
>> software RAID you can bring the drive(s) up on another box, using
>> entirely different hardware and often a different OS rev, and get to
>> the data. Both have their issues, but it's wrong to say one is 'better'
>> than
>> the other.
>>
>> -Bill Kearney

Bill Kearney <> wrote:
>> You have valid points, but I still say hardware RAID is the way to
>> go. I don't need to take a drive off one server and mount it in
>> another. I've got good backups, I've got as much redundant hardware
>> in the server as I can throw at it, and I use global hotspares so I
>> can handle two drive failures without the machine puking up blood.
>
> Well, when faced with disaster on a critical box what would you
> rather do, reload from backups or bring up the data and applications
> on another machine? What's going to best serve your user population?
> I've worked with a whole range of redundant boxes in my time and few
> ever lived up to their claims. Most failures ended up crashing the
> boxes requiring downtime anyway. Oh but if only truly standalone
> RAID (or SAN) gear wasn't so hideously expensive...
>
> No argument that backups aren't important. But given the reload time
> from most offline media that's not always as viable an option.
>
> Having a disaster recovery plan is important, especially one with
> timeframes laid out for dealing with reinstallation of the OS, apps
> and reloading of data from backups. Likewise important is having a
> plan for being able to bring a temporary box into action to serve as
> a standby during the repair process. This can actually complicate
> the process because of inconsistencies in the data added/changed
> during the use of standby gear. But if your plan includes keeping
> services active as a priority then it's just one more element to
> factor into the recovery equation.
> -Bill Kearney

My plan would involve ShadowProtect and continuous incremental backups to
another box on the network. That works really nicely.

No, no, no. The question or the issue is simple.
Can there be any factor when you Mirror two disks from Windows, why would
The Mirror disks overwhelming fail vs. the Original Disk
Example: yes, it makes sense, for each write on the Original Disk, Windows
makes 200 writes on the Mirror Disk (making this up to make a point).

Of course, the spare disks are the same lot, identical brand and model.
And, of course, they reside within the same Dell 2800 box
C:\ sits in bay 0, Mirror of C is in bay 6
E:\ sits in bay 4, Mirror of E is in bay 7

I remember in one case, the Mirror of E totally died (unreadable)[Disk
#MirZ], unmirror, take out the dead one, inserted another "good" one [Named
#MirY], while it was mirroring, the Mirror totally died and the inserted a
third disk, resynched the mirror and it worked this time. After the event,
#MirZ and #MirY are totally unreadable or unformatable.

Please keep in mind that the Original C:\ and D:\ have never failed, but
their Mirrors have failed (thrashed entirely, not reusable, not formatable)
2 and 3 times during the last 2 years. Lost 5 total disks out of spare
empty ones.

Has anybody seen a trend like this with Win 2003 Server?? I guess the
answer is No.

Never mind.....will switch to SATA storage soon.


"Bill Kearney" <> wrote in message
news: t...
>
> "Travis McGee" <> wrote in message
> news:uJ5%...
>> Have a 2003 Dell Server with 8 bays
>
> What's the temperature on the drives? If they're running too hot it might
> be shortening their lifespan. That and I seem to recall the Cheetah
> drives as running pretty hot to begin with. Is there enough airflow and
> is the room itself cool enough?

"Bill Kearney" <> wrote in message
news:P4WdnYcidbraSvzXnZ2dnUVZ_h-...
> with a whole range of redundant boxes in my time and few ever lived up to
> their claims. Most failures ended up crashing the boxes requiring
> downtime anyway.

I think redundancy in any area is kind of like that. On many occasions I
get people mad at me because I tell them that true perfect redundancy is
just a fantasy. For an example in another area, people create all kinds of
mind boggling head spinning "contraptions" for redundant internet links with
two ISPs,...that fail when disaster actually does happen.

> No argument that backups aren't important. But given the reload time from
> most offline media that's not always as viable an option.

Mine take almost an entire day to create the Full Backups. So I suspect it
would take most of that time to Restore as well. Even if the Restore is a
little quicker,..there is still time lost getting "something" loaded on the
machine do that the Restore can even begin to be run. Of course this is
assuming a total disaster requiring a full restore.

--
Phillip Windell

The views expressed, are my own and not those of my employer, or Microsoft,
or anyone else associated with me, including my cats.
-----------------------------------------------------

Phillip Windell <> wrote:
> "Bill Kearney" <> wrote in message
> news:P4WdnYcidbraSvzXnZ2dnUVZ_h-...
>> with a whole range of redundant boxes in my time and few ever lived
>> up to their claims. Most failures ended up crashing the boxes
>> requiring downtime anyway.
>
> I think redundancy in any area is kind of like that. On many
> occasions I get people mad at me because I tell them that true
> perfect redundancy is just a fantasy.

That's very true. But I don't sell anything as "perfect" - just "a lot
better than not having it." For example, most of my clients have global
hotspares in their RAID arrays ...and those have saved our ___s many times.
Sometimes two disks in an array will fail - I've seen it happen.

> For an example in another
> area, people create all kinds of mind boggling head spinning
> "contraptions" for redundant internet links with two ISPs,...that
> fail when disaster actually does happen.

I don't know what kind of contraptions you mean - but a backup WAN link is
not a bad idea. I have a few clients who have them and they tend to work.
>
>> No argument that backups aren't important. But given the reload
>> time from most offline media that's not always as viable an option.
>
> Mine take almost an entire day to create the Full Backups. So I
> suspect it would take most of that time to Restore as well. Even if
> the Restore is a little quicker,..there is still time lost getting
> "something" loaded on the machine do that the Restore can even begin
> to be run. Of course this is assuming a total disaster requiring a
> full restore.

Do check out the ShadowProtect line and continuous incrementals.