Converting IDE RAID 1 to SATA RAID 1

[JonnyCab®]

Hola,

I have a server with an on-board IDE RAID controller controlling two IDE
drives in a RAID 1 (mirror) configuration.

If I install a Promise TX2300 PCI-to-SATAII RAID controller and two SATAII
drives, can I build the new SATA RAID 1 array from the existing IDE array,
or at least populate one SATA drive?

Can I then remove the IDE array, disable the IDE controller, and boot
normally from the new SATA controller?

This process would be under some time constraints if I do it, so I need it
to go just that smoothly if I decide to do it.

Thanks!

 

[Arno Wagner]

Hola,

I have a server with an on-board IDE RAID controller controlling two IDE
drives in a RAID 1 (mirror) configuration.

If I install a Promise TX2300 PCI-to-SATAII RAID controller and two SATAII
drives, can I build the new SATA RAID 1 array from the existing IDE array,
or at least populate one SATA drive?

Very unlikely. What you can do is to use ordinary mirroring software to
copy the contenst of the array. The problem is that the two RAID
controllers very likely don't understand each other and cannot have
shared arrays.

Can I then remove the IDE array, disable the IDE controller, and boot
normally from the new SATA controller?

This process would be under some time constraints if I do it, so I need it
to go just that smoothly if I decide to do it.

Then you cannot do it. You will need to look for a different option.
You very likely will have to accept the downtime mirroring the array
with external software will bring with it. Hardware/controller RAID
vendors do not want to be compatible with each other. They benefit
from you having to buy new hardware and of course once you have
a controller from one vendor they do _not_ want you to be able to
migrate away.

Side note: Your idea would work if the original and target array
was any decent software RAID, since then the controllers do not
matter. For example with Linux Software-RAID you could add the
the two SATA drives and then do the rest while the Server is running,
i.e. kick one of the original disks, add one of the SATA disks, wait
for resync to complete, kick the second original disk and add the
second SATA disk. I have done things like this without trouble.

Side note 2: Have you asked yourself what you will do if your
mainboard fails? You might need the same mainboard for fast
access to your data. Same goes for an external RAID controller.
You might have to buy two to be reasonably safe.

Arno

 

[Folkert Rienstra]

Very unlikely. What you can do is to use ordinary mirroring software to
copy the contenst of the array. The problem is that the two RAID
controllers very likely don't understand each other and cannot have
shared arrays.



Then you cannot do it.

Nonsense. Ever heared of backups? Or just cloning?

You will need to look for a different option.

You very likely will have to accept the downtime mirroring the array
with external software will bring with it.

Whatever that was meant to read.

Hardware/controller RAID
vendors do not want to be compatible with each other. They benefit
from you having to buy new hardware and of course once you have
a controller from one vendor they do _not_ want you to be able to
migrate away.

Side note: Your idea would work if the original and target array
was any decent software RAID, since then the controllers do not
matter.

They never matter if you have backups. And since they aren't
software raid they behave like single drives and any decent
but simple cloning software or backup sofware should work.

 

[JonnyCab®]

How about this?

* Power off
* Remove one IDE drive from the array
* Connect one of the new SATA drives with an IDE adapter
* Power on
* Rebuild the array (populating the new SATA drive)
* Power off
* Install both SATA drives and the SATA RAID controller
* Power on
* In BIOS, disable the IDE RAID controller, save and reboot
* Populate the second SATA drive

Done?

My experience has been rebuilding an array takes a lot less time than using
software.

Am I wrong about any of this?

Thanks again!

 

[Folkert Rienstra]

How about this?

* Power off
* Remove one IDE drive from the array
* Connect one of the new SATA drives with an IDE adapter
* Power on
* Rebuild the array (populating the new SATA drive)
* Power off
* Install both SATA drives and the SATA RAID controller
* Power on
* In BIOS, disable the IDE RAID controller, save and reboot
* Populate the second SATA drive

Done?

Sure, but all data was transferred twice.

My experience has been rebuilding an array takes a lot less time than using
software.

Am I wrong about any of this?

Should not make much difference compared with a sector cloner or
with a physical image backup. And you do that only once, no rebuild.

Copying by filesystem usually has the big penalty.

 

[JonnyCab®]

Should not make much difference compared with a sector cloner or

with a physical image backup. And you do that only once, no rebuild.

Copying by filesystem usually has the big penalty.

Do you mean by letting the array do the rebuild, or software? My experience
has been that letting the array build the second drive takes FAR less time
than letting something like Ghost do it. Maybe it's just Ghost...dunno...

Do you have a favorite program that would change my mind?

 

[Arno Wagner]

How about this?

* Power off
* Remove one IDE drive from the array
* Connect one of the new SATA drives with an IDE adapter
* Power on
* Rebuild the array (populating the new SATA drive)
* Power off
* Install both SATA drives and the SATA RAID controller

Unclear whether this works. Unlikely IMO. This will only work
if both controllers use compatible metadata for the array
description. It should work if both controllers operate
metadata-free, i.e. do not store anything on the disks
except the data. But usually a RAID controller stores a
descriptor of the array on each disk in some non-standardised
location and format.

* Power on
* In BIOS, disable the IDE RAID controller, save and reboot
* Populate the second SATA drive

My experience has been rebuilding an array takes a lot less time than using
software.

Am I wrong about any of this?

Yes, I think so. See above.

Arno

 

[JonnyCab®]

How about this?

Unclear whether this works. Unlikely IMO. This will only work
if both controllers use compatible metadata for the array
description. It should work if both controllers operate
metadata-free, i.e. do not store anything on the disks
except the data. But usually a RAID controller stores a
descriptor of the array on each disk in some non-standardised
location and format.

I do see your point. Wouldn't the newer SATA controller somehow see the new
disk and "adopt" it as its own (for the lack of a better way of describing
the process)? I converted two machines from single IDE to RAID 1 by simply
installing the original single drive into a new RAID 1 IDE array, building
the array with one new disk, then removing the original disk and installing
a second new disk, and then having the first disk build the second. It
worked perfectly, and took only a few hours.

Here are some details: The servers (I actually want to do this to two of
them) are Tyan Tiger 200T (S2505T) boards with Promise FastTrak100 RAID IDE
controllers. They've worked rock-solid flawlessly in RAID 1 configurations
for several years, and I want to continue to use them until I can upgrade
the boards, processors, power supplies, and memory, probably in a year or
two. I replace the drives (as preventative maintenance) each year. For
now, I want to move the IDE drives out of the inside of the cases and into
fan-ventilated I-Star BPU-230 SATA enclosures, so the drives can be replaced
when necessary without taking the cases apart . Since I'm over 500 miles
away from the servers, it's much easier for me to instruct someone to
replace a drive in a tray than inside a cage filled with cables, especially
when the case has to be pulled out and disassembled.

Anyway, the new controller will be a Promise TX2300 with a pair of Maxtor
7L250S0 Maxline III 16MB 250GB (5-year, enterprise-level) drives.

I'm not a fan of Ghost, which is the only "cloning" software I've used.
I've seen it take almost a day to run, and I don't have that kind of time to
wait for it. I've put new drives into RAID arrays and watched the array
take under two hours to copy to the new disk, so you can see why I'm a
little hesitant to use software.

Thanks for any additional insight, and/or any suggestions for software that
doesn't run like it's written by Microsoft or Symantec. LOL

 

[Folkert Rienstra]

Do you mean by letting the array do the rebuild, or software?

No rebuild at all. Just copying data to the new empty RAID drive.

My experience has been that letting the array build the second drive takes FAR less time

No rebuild necessary.

than letting something like Ghost do it. Maybe it's just Ghost...dunno...

Must be.

Do you have a favorite program that would change my mind?

The most simple drive cloner that you can find.
If a DOS one, set-up the disk cache program first.

 

[Arno Wagner]

I do see your point. Wouldn't the newer SATA controller somehow see the new
disk and "adopt" it as its own (for the lack of a better way of describing
the process)?

No. If it is connected to the IDE controller no other hardware has
any business messing with it. It likely cannot even see it without
sort of "hacking" the OS driver layer.

I converted two machines from single IDE to RAID 1 by simply
installing the original single drive into a new RAID 1 IDE array, building
the array with one new disk, then removing the original disk and installing
a second new disk, and then having the first disk build the second. It
worked perfectly, and took only a few hours.

Ok. So the old controller may not store a superblock on disk. Or your
disks are short one sector and you are lucky it was not allocated
(smart controllers store it at the end of the disk and some partition
tools dont use the last few MBs on disk). You might also have one
sector you better not write any data to.

Here are some details: The servers (I actually want to do this to two of
them) are Tyan Tiger 200T (S2505T) boards with Promise FastTrak100 RAID IDE
controllers. They've worked rock-solid flawlessly in RAID 1 configurations
for several years, and I want to continue to use them until I can upgrade
the boards, processors, power supplies, and memory, probably in a year or
two. I replace the drives (as preventative maintenance) each year. For
now, I want to move the IDE drives out of the inside of the cases and into
fan-ventilated I-Star BPU-230 SATA enclosures, so the drives can be replaced
when necessary without taking the cases apart . Since I'm over 500 miles
away from the servers, it's much easier for me to instruct someone to
replace a drive in a tray than inside a cage filled with cables, especially
when the case has to be pulled out and disassembled.

Anyway, the new controller will be a Promise TX2300 with a pair of Maxtor
7L250S0 Maxline III 16MB 250GB (5-year, enterprise-level) drives.

I'm not a fan of Ghost, which is the only "cloning" software I've used.
I've seen it take almost a day to run, and I don't have that kind of time to
wait for it. I've put new drives into RAID arrays and watched the array
take under two hours to copy to the new disk, so you can see why I'm a
little hesitant to use software.

Thanks for any additional insight, and/or any suggestions for software that
doesn't run like it's written by Microsoft or Symantec. LOL

Depending on what your server does, you can make an image or backup
while the server runs and recreate the installation on a separate
system on the SATA disks and then swap them in. You need to have
some sensible backup procedure anysways that lets you do something
close.

Probably the easiest approach for now is to stay with IDE and just move
the disks into IDE enclosures that can be removed from the front.

Arno

 

[JonnyCab®]

No rebuild at all. Just copying data to the new empty RAID drive.

Would the new drive be bootable then (from the SATA RAID controller), once I
disable the IDE controller?

 

[JonnyCab®]

I do see your point. Wouldn't the newer SATA controller somehow see the
new

No. If it is connected to the IDE controller no other hardware has
any business messing with it. It likely cannot even see it without
sort of "hacking" the OS driver layer.

Hmmm...I meant to ask if the SATA controller would be able to "see" (and
boot from) the disk created by the IDE controller once it was removed from
the IDE controller and connected as a single drive to the SATA controller.
I was not going to enable both controllers at the same time.

Ok. So the old controller may not store a superblock on disk. Or your
disks are short one sector and you are lucky it was not allocated
(smart controllers store it at the end of the disk and some partition
tools dont use the last few MBs on disk). You might also have one
sector you better not write any data to.

Well, it's worked fine for almost two years. I'd have to say I trust
Promise pretty well by now, so that sector may be protected by the
controller.

Depending on what your server does, you can make an image or backup
while the server runs and recreate the installation on a separate
system on the SATA disks and then swap them in. You need to have
some sensible backup procedure anysways that lets you do something
close.

Probably the easiest approach for now is to stay with IDE and just move
the disks into IDE enclosures that can be removed from the front.

I want to use the SATA enclosures because of the 80mm, long-life fans, not
the cheesy 40mm sleeve-bearing fans that are always installed in the
individual enclosures. Those fans last for less than a year, they don't
warn of their failure, and they leave the drive alone to fry itself. I also
wanted to take advantage of SATA's slightly faster speeds and NCQ, without
having to spend more $$$ on updating the board/processor/memory. I have to
work withing a company budget right now.

 

[Folkert Rienstra]

No. If it is connected to the IDE controller no other hardware has
any business messing with it. It likely cannot even see it without
sort of "hacking" the OS driver layer.

So that's obviously not what he meant.

That's what he meant, obviously.

Ok.

Sofar for your "Unlikely" 'opinion'.

So the old controller may not store a superblock on disk.

Or doesn't care that the imported drive has one.

Or your disks are short one sector and you are lucky it was not allocated

Or maybe he just uses a capable controller.

(smart controllers store it at the end of the disk and some partition
tools dont use the last few MBs on disk).

And a smart controller would check and protect that area.

You might also have one sector you better not write any data to.

As if a ('smart') Raid controller would allow that.

What about your SATAs then?

Depending on what your server does, you can make an image or backup
while the server runs and recreate the installation on a separate system
on the SATA disks and then swap them in. You need to have some sensible
backup procedure anysways that lets you do something close.

Probably the easiest approach for now is to stay with IDE and just move
the disks into IDE enclosures that can be removed from the front.

Probably the easiest approach for now is to ignore you completely since
you obviously have even less experience on the subject than OP.

 

[Folkert Rienstra]

Would the new drive be bootable then (from the SATA RAID controller), once I
disable the IDE controller?

Should be, if a physical clone. Otherwise it wouldn't be a 'clone'.

 

[JonnyCab®]

Er, make that *within* a company budget. LOL

 

[Arno Wagner]

Hmmm...I meant to ask if the SATA controller would be able to "see" (and
boot from) the disk created by the IDE controller once it was removed from
the IDE controller and connected as a single drive to the SATA controller.
I was not going to enable both controllers at the same time.

I see. Depends on whether they both use metadata or not
and if they do whether it is identical.

Well, it's worked fine for almost two years. I'd have to say I trust
Promise pretty well by now, so that sector may be protected by the
controller.

Most likely it is. I agree.

I want to use the SATA enclosures because of the 80mm, long-life fans, not
the cheesy 40mm sleeve-bearing fans that are always installed in the
individual enclosures. Those fans last for less than a year, they don't
warn of their failure, and they leave the drive alone to fry itself. I also
wanted to take advantage of SATA's slightly faster speeds and NCQ, without
having to spend more $$$ on updating the board/processor/memory. I have to
work withing a company budget right now.

Well, yes. I can understand your reasoning. But I fear you will
at least have to experiment and you may discover that what you want
to do is not that easy.

Arno

 

[JonnyCab®]

No rebuild at all. Just copying data to the new empty RAID drive.

Should be, if a physical clone. Otherwise it wouldn't be a 'clone'.

I was just wondering, since I've cloned entire bootable drives before (going
from one interface to another), and the cloned drive was not always
bootable.

I'm going to try to have the IDE controller build one of the SATA drives,
then install it on the SATA controller to see if it boots with the IDE
controller disabled. It's worth a try...

Thanks again.

 

[=?iso-8859-1?B?Sm9ubnlDYWKu?=]

Here's a quick update:

* I installed the new controller with the new disks
* I configured the new array as RAID 1 (mirror) when it booted
* I booted using Maxtor's MaxBlast 4 floppy
* I cloned the original IDE RAID drive onto the SATA RAID drive
* I disconnnected the IDE RAID controller and tried to boot from the
new controller

As I expected, I got the blue-screen inaccessible boot device message.

So, I booted into the original Win2K installation to install the
drivers for the TX2200 card, and now I'm cloning the IDE array to the
SATA array again.

Wanna take bets on whether or not the SATA array will boot this time?

And so much for my prior experience with cloning software. MaxBlast
cloned the IDE array (80GB drives about 3/4 full) to the SATA array
(250GB drives) in under half an hour. At least this doesn't take too
much time away from experimenting. LOL

 

[=?iso-8859-1?B?Sm9ubnlDYWKu?=]

PS...I'm not wild about doing a Win2K repair install, but at least the
F6 option would allow me to use the new array's drivers and boot from
that array. Then I can do a few hours worth of Windows Updates. <gak>

That would be a last resort.

 

[Folkert Rienstra]

Here's a quick update:

* I installed the new controller with the new disks
* I configured the new array as RAID 1 (mirror) when it booted
* I booted using Maxtor's MaxBlast 4 floppy
* I cloned the original IDE RAID drive onto the SATA RAID drive
* I disconnnected the IDE RAID controller and tried to boot from the
new controller

As I expected, I got the blue-screen inaccessible boot device message.

So, I booted into the original Win2K installation to install the
drivers for the TX2200 card,

and now I'm cloning the IDE array to the SATA array again.

That is *so* braindead.