Windows 7 Disk Management: spanned volume won't re-integrate missingdisk

[miso]

I agree. I just advise toget that spare controller and make
sure swapping it out works as expected.


Ah, sorry. What I meant is that you are hardware dependent in
the sense that you are dependent on this specific controller
type. As long as you have a spare, you can recover the array
from a broken controller on any other hardware. But well
implemented software RAID is not slower or less reliable and
does away with the need for any spare hardware. If you have
that spare controller, (and it being driverless is definitely
a huge advantage!), the difference is small. If you do not
have it and then find out that it is out of production or
otherwise hard to get when your main controller fails, the
difference is huge...


Indexing is another of these broken "features". On my laptop
it made everything choppy until I turned it off. Never
needed it anyways, I know where I keep stuff.

Arno

In theory, these RAID volumes should follow a standard. That doesn't
seem to be the case in practice. I still have a RAID 5 array stashed
away that I'm going to try to recover some day. I had a mobo fail and it
used a fake raid. The RAID10 could be recovered in another PC without
issue, but not Raid 5. I tried to recover it with Pared magic, but it
just couldn't put the array back together I had a hackup, but not
totally up to date.

There are FreeNAS proponents that prefer to do the RAID completely in
the OS.

 

[Char Jackson]

Wow! You have *big* systems. Are these home systems or are they used
in a business somewhere?

Home systems.

Both systems are just about out of capacity so I'm watching HDD prices with
the idea that the 15-drive system (16 drives if you include the OS) can be
expanded to 24 drives and the 9-drive system (10 with the OS) can be
expanded to 13 drives. By then, both systems will be physically out of room
and it's either time for an additional server or piecemeal replacement of
2TB drives with bigger units.

 

[Char Jackson]

Are these all simple spans, or does this software also do software
RAID-5 or other RAID's?

Drivebender is simple spanning, (they call it pooling, as in older versions
of Windows Home Server). You can designate any number of folders for
duplication and Drivebender will ensure that the two copies are always on
different physical disks, but if you need more than that you have to add it
yourself. SnapRAID or FlexRAID make nice additions, for example.

 

[Arno]

In theory, these RAID volumes should follow a standard. That doesn't
seem to be the case in practice.

Sort-of but not quite. The on-disk format is sort-of standardized,
which basically just stripe size and order variable. But the
metadata (disk ID, array ID, flags for all kinds of things, etc.)
is not at all standardized.

I still have a RAID 5 array stashed
away that I'm going to try to recover some day. I had a mobo fail and it
used a fake raid. The RAID10 could be recovered in another PC without
issue, but not Raid 5. I tried to recover it with Pared magic, but it
just couldn't put the array back together I had a hackup, but not
totally up to date.

There are FreeNAS proponents that prefer to do the RAID completely in
the OS.

I have been using Linux software raid for 10 years without problems,
almost all running 24/7 and one installation was the two fileservers
for a computing cluster.

Arno

 

[Yousuf Khan]

I suspect that the user accepted Windows' invitation to initialise the
drive. Many users do this without realising that "initialisation" is
data destructive, probably because the term sounds innocuous.

That's a possibility, but he didn't mention doing that, but I'll ask him
directly. He did mention trying the "Reactivate" command in Disk
Management, which of course didn't work. Could Reactivate call an
Initialize? Judging by how quickly & automatically things came right
back after the reboot when the partition table was intact, I would've
expected no need to even touch any of these commands, so either they got
run before the reboot (in a desperate attempt to get things back) and
ended up causing the problem after a reboot, or something truly external
erased the partition table perhaps during the communications problem to
that disk.

Yousuf Khan

 

[Ken Blake]

Home systems.

Both systems are just about out of capacity so I'm watching HDD prices with
the idea that the 15-drive system (16 drives if you include the OS) can be
expanded to 24 drives and the 9-drive system (10 with the OS) can be
expanded to 13 drives. By then, both systems will be physically out of room
and it's either time for an additional server or piecemeal replacement of
2TB drives with bigger units.

Wow, again! You must have lots of very big files--video files perhaps.

If you add up everything on all three of my drives here, it comes to
something around half a terabyte. Add in the drives in the other three
systems in my home--my wife's machine, my laptop, and my Windows Home
Server--and the total is around 1.5 TB.

 

[Franc Zabkar]

He did mention trying the "Reactivate" command in Disk
Management, which of course didn't work. Could Reactivate call an
Initialize?

I don't know, but I would think not. Microsoft's technical articles
aren't really helpful in this regard, though. I like to see what
happens at the bits-and-bytes level, but Microsoft rarely goes that
deep.

- Franc Zabkar

 

[Paul]

I don't know, but I would think not. Microsoft's technical articles
aren't really helpful in this regard, though. I like to see what
happens at the bits-and-bytes level, but Microsoft rarely goes that
deep.

- Franc Zabkar

Reactivate might exist for the purpose of handling a "hot inserted"
span or RAID member. Like plugging a SATA drive into a SATA backplane
with the power on.

If you're cold-booting with all members restored
in a set, it likely puts the mess back online all by itself.

There's no reason for "Reactivate" to delete an MBR. The
only thing that would do that, might be "convert Dynamic to Basic",
complete with its own warning dialogs etc. As a Dynamic to Basic
conversion is going to remove the protective 0x42 value and
allow the MBR to hold regular partition type entries.

http://www.win.tue.nl/~aeb/partitions/partition_types-1.html

"42 Windows 2000 dynamic extended partition marker"

Paul

 

[Arno]

A great many Linux, BSD and Unix systems use software raid - there are
many reasons why it is often preferably to hardware raid. And pretty
much every low or medium price NAS system you can buy uses either Linux
or BSD with software raid.

In the windows world, fake raid or hardware raid is the norm, but in the
*nix world software raid is very common.

I was wondering about the reasons for a long time. By now
I have come to the conclusion that the typical Unix Admin
is expected to be able to handle software RAID configuration,
while the typical Windows admin is at best expected to be
able to plug in disks.

Arno

 

[Yousuf Khan]

Reactivate might exist for the purpose of handling a "hot inserted"
span or RAID member. Like plugging a SATA drive into a SATA backplane
with the power on.

The Windows 7 version (even for Ultimate) of Disk Management doesn't
include any RAID beyond spanning, striping, or mirroring. I found that
out the hard way, when I tried to create some RAID5 partitions, and
found it all greyed out. Looked it up and saw it was only for Windows
Server. I'm not sure I'd trust the Disk Management on Server editions
either, considering the stuff we went through here.

If you're cold-booting with all members restored
in a set, it likely puts the mess back online all by itself.

I think this might have contributed to the mess here. He knew enough to
reboot the machine, but he didn't consider doing a complete shutdown and
restart to re-initialize the peripherals and power supply too. He was
simply doing warm reboots, and it wasn't fixing anything. Then
eventually panic and desperation set in, and he may have done something
without completely realizing the implications.

However, I might be completely off-base in suspecting human error here,
I did ask him today if he used the "Initialize" command, and watched him
recreating his steps, he didn't even know where to find it, so it's not
likely he hit it by accident.

There's no reason for "Reactivate" to delete an MBR. The
only thing that would do that, might be "convert Dynamic to Basic",
complete with its own warning dialogs etc. As a Dynamic to Basic
conversion is going to remove the protective 0x42 value and
allow the MBR to hold regular partition type entries.

Now, let's take a hypothetical situation where the Windows 7 version of
Disk Management *does* support software RAID5, or maybe even just
mirroring. If a disk dies and you replace it with a new disk, would the
Reactivate be used to resync the new drive to the existing volume? If
that's the case, then maybe it would have to call "Init" to setup the
new drive as a dynamic drive?

Yousuf Khan

 

[Yousuf Khan]

A great many Linux, BSD and Unix systems use software raid - there are
many reasons why it is often preferably to hardware raid. And pretty
much every low or medium price NAS system you can buy uses either Linux
or BSD with software raid.

In the windows world, fake raid or hardware raid is the norm, but in the
*nix world software raid is very common.

Somebody ought to port the Linux md raid to Windows one day. I doubt it
would be nearly as capable as it is on Linux, because they won't have
access to some of the low-level information about Windows to allow it to
work with Windows boot disks, but at least it'll work with secondary
data disks.

Yousuf Khan

 

[DevilsPGD]

In the last episode of <[email protected]>,

Home systems.

Both systems are just about out of capacity so I'm watching HDD prices with
the idea that the 15-drive system (16 drives if you include the OS) can be
expanded to 24 drives and the 9-drive system (10 with the OS) can be
expanded to 13 drives. By then, both systems will be physically out of room
and it's either time for an additional server or piecemeal replacement of
2TB drives with bigger units.

What sort of cases/enclosures are you using? I've got a couple boxes
that are reaching their limits due to physical room in the case and I'm
eyeballing replacements but haven't found anything that isn't horribly
over-priced yet.

My goal is 12-16 SATA drives per case, although +2 would be ideal so
that I can have a couple drives for the OS and fully fill my 2x8 SATA
RAID cards to capacity.

I was looking at
http://www.newegg.ca/Product/Product.aspx?Item=N82E16811147165 and
actually had one on order, but I needed a pair and they got discontinued
after I ordered one, so I cancelled the order.

 

[Yousuf Khan]

How is it that Disks 0 and 2 appear to have swapped places? Does it
matter?

Yeah, sorry that was another complication, this friend had just recently
installed an SSD, and we were dual-booting between it and the old HDD
boot, from time to time. So depending on whether we were booted through
the HDD or the SSD, the disk numbers changed position. I listed the
original disk listing when booted through the HDD, but I didn't feel
like waiting around for an HDD boot everytime, so began working through
the SSD instead. But don't worry, I kept the disks straight in my head.

In fact he was a little concerned that the SSD had something to do with
this span disk failure, because it happened so soon after installation
of the SSD. But I told him they're not related. However, I do think that
maybe the addition of the SSD might have been related through power
supply sharing issues, which I didn't tell him about.

Yousuf Khan

 

[Char Jackson]

In the last episode of <[email protected]>,


What sort of cases/enclosures are you using? I've got a couple boxes
that are reaching their limits due to physical room in the case and I'm
eyeballing replacements but haven't found anything that isn't horribly
over-priced yet.

One is a Norco RPC-450B, advertised as being able to hold just 10 drives,
but the standard trick is to remove the optical drive bay and install a
third HDD bay, increasing its internal capacity to 15 3.5" drives plus a
2.5" drive mounted in a card slot on the mobo.

<http://www.newegg.com/Product/Product.aspx?Item=N82E16811219030>

At some point, I intend to replace the 450B (or build another system
entirely) with a Norco 24-drive chassis such as the RPC-4220 or 4224.

<http://www.newegg.com/Norco-Technologies-Inc/BrandStore/ID-10473>

My other case is a Lian-Li tower case, not sure of the model, that natively
holds 8 3.5" drives, but with creative use of adapters I currently have 10
drives installed and see a viable way to get to at least 13 or 14.

My goal is 12-16 SATA drives per case, although +2 would be ideal so
that I can have a couple drives for the OS and fully fill my 2x8 SATA
RAID cards to capacity.

I was looking at
http://www.newegg.ca/Product/Product.aspx?Item=N82E16811147165 and
actually had one on order, but I needed a pair and they got discontinued
after I ordered one, so I cancelled the order.

From a distance, I could confuse that Rosewill with my Norco.

 

[DevilsPGD]

In the last episode of <[email protected]>,

My other case is a Lian-Li tower case, not sure of the model, that natively
holds 8 3.5" drives, but with creative use of adapters I currently have 10
drives installed and see a viable way to get to at least 13 or 14.

Cool, thanks. I don't mind using adapters to turn external bays into
3.5" bays, but I'd love to find something that doesn't need it.

From a distance, I could confuse that Rosewill with my Norco.

Yeah. They have a non-hotswappable one that is even closer, but I do
like the idea of hotswappable bays since my RAID cards are all hotswap
capable/ready at this point.

OTOH, I guess there's not really a lot of space for innovation when it
comes to packing a lot of drives in a case with a motherboard at the
rear.

 

[Char Jackson]

In the last episode of <[email protected]>,


Yeah. They have a non-hotswappable one that is even closer, but I do
like the idea of hotswappable bays since my RAID cards are all hotswap
capable/ready at this point.

Some of the Norco cases are hot swappable, but I use Drivebender as a
pooling service and it gets testy when a pooled volume disappears.

OTOH, I guess there's not really a lot of space for innovation when it
comes to packing a lot of drives in a case with a motherboard at the
rear.

IIRC, Google just says to heck with the case and the AC power supply,
letting the mobo, drives, and DC supply just lie there naked. I suppose that
could be an option, but I like the finished look of having that stuff in a
case.

 

[DevilsPGD]

In the last episode of <[email protected]>,

IIRC, Google just says to heck with the case and the AC power supply,
letting the mobo, drives, and DC supply just lie there naked. I suppose that
could be an option, but I like the finished look of having that stuff in a
case.

That's not an option where I live, at least for the most part. We have
one machine mounted in a cabinet behind the TV, but other than that,
it's not practical.

 

[Franc Zabkar]

Okay, PTEdit worked!!! The partition table on this disk was showing Type
0 "empty", so I used Ptedit to turn them into Type 42 (Dynamic Disk). I
also had to fill in the remaining fields in that partition table by
hand: Starting Cylinder/Head/Sector = all zeros; Ending Cyl/Head/Sector
= 1023/254/63; Sectors Before = zero; Sectors End = Total Sectors -
2111. I found these other parameters out by comparing them to the other
disks in the dynamic volume.

I confess that I know nothing about dynamic disks but ISTM that the
CHS numbers don't make sense.

Firstly, sector numbers begin counting from 1, not 0.

C/H/S values of 1023/254/63 would suggest that the partition size is
about 16 million sectors, not 2111. I realise that these values are
used when the partition size is greater than 8GB, but that doesn't
seem to fit with the rest of the partition table data.

Furthermore, 2111 = 63 + 2048, so I'm wondering whether there is a
1MiB partition beginning at sector 63. Sector 63 is where the first
traditional MBR partition begins, not LBA 0.

In fact the following URL has this to say:

"If a partition table entry of type 0x42 is present in the legacy
partition table, then W2K ignores the legacy partition table and uses
a proprietary partition table and a proprietary partitioning scheme
(LDM or DDM). As the Microsoft KnowledgeBase writes: Pure dynamic
disks (those not containing any hard-linked partitions) have only a
single partition table entry (type 42) to define the entire disk.
Dynamic disks store their volume configuration in a database located
in a 1-MB private region at the end of each dynamic disk."

http://www.win.tue.nl/~aeb/partitions/partition_types-1.html

- Franc Zabkar

 

[Yousuf Khan]

I know that XP Pro only supports RAID0 and RAID1 out of the box, and you
need a server version to support RAID5. But support for RAID5 can be
"added" to XP by a few registry tweaks. Maybe the same applies with
Win7? Of course, such tweaks are unlikely to be supported by MS...

Possibly, however this is Microsoft's official response to why RAID-5 is
greyed out in Disk Management:

"New RAID-5 Volume..." greyed out in Disk Management
http://social.technet.microsoft.com...l/thread/a1851e42-c705-4558-920c-30ba7c6cf080

So I guess RAID-5 requires even more expenditure to get enabled, beyond
Windows 7 Ultimate. You need to go straight to Windows Server.

But the following thread even mentions that even if I had Windows
Server, I probably wouldn't really want to use the software RAID-5 on
it, because it's so slow:

Establishing a software RAID5 with Diskpart seems to take DAYS with no
end in sight - what gives?
http://social.technet.microsoft.com...s/thread/df7228a9-98d3-431a-b530-f513b6141608

Yousuf Khan

 

[Yousuf Khan]

I confess that I know nothing about dynamic disks but ISTM that the
CHS numbers don't make sense.

Firstly, sector numbers begin counting from 1, not 0.

Yup, sorry, I made a mistake when writing that one, I was operating from
memory since I wasn't near the original system. The actual values for
that should be:

Starting Cylinder/Head/Sector = 0/1/1

The Ending CHS values are right though.

C/H/S values of 1023/254/63 would suggest that the partition size is
about 16 million sectors, not 2111. I realise that these values are
used when the partition size is greater than 8GB, but that doesn't
seem to fit with the rest of the partition table data.

No, I'm not saying that the partition size is only 2111 sectors, I'm
saying that you need to _subtract_ 2111 sectors from the total number of
sectors and put that value in here, let me rewrite it on its own line, thus:

Sectors End = Total Sectors - 2111

So if you have 1,000,000 sectors total in your disk, then you would put
997,889 sectors in this field, i.e. 1000000 - 2111 = 997889.

2111 sectors is about equal to 1 MB, which would be just about the right
size for the metadata database for the dynamic disks.

Furthermore, 2111 = 63 + 2048, so I'm wondering whether there is a
1MiB partition beginning at sector 63. Sector 63 is where the first
traditional MBR partition begins, not LBA 0.

Probably nearly right, except I'm thinking that the 63 + 2048 is around
the end of the disk, rather than around the beginning.

Yousuf Khan