Moving 2003 to new disks scares me.

[Deepseadata]

I thought I'd throw this post up while I do my research.

I've got SBS2003 on a Dell Power Edge 2950 (with 3 x 146GB SAS SCSI's)running as my only domain controller and Exchange Server.

I have no space left on any of my disk drives. And Exchange dismounting everyday. I have read that I can increase my Exchange store up to 75gb but the drive is physically full!

When I arrived here as admin I noticed that nothing was mirrored or RAIDed. They have put the OS on C:, Userdata on D:, Spare: E (which is full - LOL).

I have just bought 4 2TB SCSI drives and I'm scratching my head trying to figure out how to put them into action.

I have a spare 2950 chassis and was thinking of using 2 spare 146gb's to mirror and fresh install of 2003. Then I could use the 2TB's for a RAID5 for all my data.

Then I have to figure out how to get all my databases and shares working properly again. This scares me.

How would you do it?

 

[Roadki11]

Here is what I would do. I would attach a large IDE or SCSI drive in the server. I would then ghost the 3 drives to images on to the added storage. Then pull the 3 146gig drives and mark them as to which slot they came from just in case you need to revert back if something goes wrong. Install the new drives and set up a raid5 or 6 (raid6 will require more drives). Bear in mind a 2TB drive is going to take forever to recover because the disk so large if a drive fails, just an added point. You could do a 3 disk raid5 with a hot spare if the raid card allows it. You would still have 4TB of storage and some added fault tolerance. Then restore the images back to your new raid. You can keep the C:\, D:\ and E:\ layout and can increase the partition size during the restore. If you do a 4 disk raid 5 I would get a spare drive, you don’t want to wait around for a drive to ship if you have a disk failure at some point.

Lots of ways to handle this, I am sure you will get other opinions.


RoadKi11

"This apparent fear reaction is typical, rather than try to solve technical problems technically, policy solutions are often chosen." - Fred Cohen

 

[itsp1965]

One thing you may need to take into consideration is that the newer larger drives support a 4K sector size which may be incompatible with the way your existing OS disks are formatted (using standard 512K sectors). Since you say you have access to a secondary 2950 chassis you may want to test restoring those images on the secondary machine

 

[Deepseadata]

Thanks to you both. These are very good suggestions.

I fired up my spare yesterday and was thinking of starting a fresh 2003 install from scratch. I didn't have enough drive caddies around though. I ordered some from ebay and hope they arrive soon.

Ideally, I could migrate the databases from the old drive to the new install with some research.

Then I can slowly copy the settings over to the new install and hopefully put the spare into service!

 

[goombawaho]

I would agree with Roadki11's idea and that's probably what I would have suggested. This is an excellent way to go because you won't be putting any of your data at risk by trying/doing it this way. UNLESS you ghost the wrong way - just be careful of that. Done it - don't want to do it again.

I would tell you that the last thing you do before shutting down the server before ghosting is this. Do a CHKDSK on the D: and E: (one at at time), then schedule a disk check of C: and shut down, start it up and and allow the CHKDSK to run. Then shut down (properly) and get your drives ghosted.

The only thing that might screw you up is if there are bad sectors and then you might need to use the "ignore bad sector" switch in Ghost, which is an option. But try it first without that switch.

Especially important points made above:
"mark them as to which slot they came from just in case you need to revert back if something goes wrong"

"increase the partition size during the restore" in Ghost before saying "OK" to the image restore to the new drives.

Also - name your ghost images so that there is no confusion as to which it is: Drive1.GHO, Drive2.GHO, etc.

 

[technome]

Agree with Roadkill and Goombawaho, this is the way to go. Running chkdsk is critical unless you enjoy seeing a possible failure of the imaging to the new array.

Be aware raid 5 was not designed with very large capacity disks in mind. If multiple physical errors build up before the controller can mark them as "bad", the raid can failed, most likely during a rebuild. Raid 6 would be a better choice as concerning safety.

If you go with raid 5, ensure "Patrol Reads" (in raid controller parameters) is ALWAYS enabled. "Patrol reads" tests every sector of the array drives, unlike a "Consistency Check", which only checks sectors of the disk platters which contain data. With large disks, array failures are caused primarily by errors building up in sectors NOT occupied by data; the controller fails the array as to many errors are found at one time, as in a rebuild. Patrol reads correct the physical errors before they have a chance to build up to a level the controller can't handle at one time.



........................................
Chernobyl disaster..a must see pictorial

http://www.kiddofspeed.com/default.htm

 

[goombawaho]

Be aware raid 5 was not designed with very large capacity disks in mind. If multiple physical errors build up before the controller can mark them as "bad", the raid can failed, most likely during a rebuild.

^^^^^^^^^^^^^^^
I don't necessarily agree/disagree with this statement, but I've never seen warnings or limitations about RAID5 and individual disk size. It's always been about what the controller can recognize.

I believe that the bad block situation is handled by TLER and is not a problem if you have the correct type of drives in the RAID.

http://wdc.custhelp.com/app/answers...-edition-and-raid-(enterprise)-edition-drives

 

[technome]

This link goes into the issue a bit, could not find an Intel paper I was looking for....

http://blog.sonic-lux.net/wp-content/uploads/2009/08/raid6_2.pdf

"I believe that the bad block situation is handled by TLER and is not a problem if you have the correct type of drives in the RAID"

Yes, the average raid controller can handle a few errors during a rebuild or initalization, but with larger disks or large capacity arrays, if the number of error is more then a few (easily can happen with large disks/arrays) the array fails; with small disks or small array capacity this is very unlikely to happen. With Enterprise disk subsystems, the systems we could not afford, the disk controller are far more robust and can handle all the errors which could found. What I do understand is why the controllers have not kept up in the error handling capacity arena, inline with present capacities.
For that matter I never understood why controllers do not offer the ability to continue, with warnings, when errors exceed the controller's ability for correction...I would rather have some a few corrupt files then a failed array containing a million files.

calculator of failure....

http://www.raid-failure.com/raid5-failure.aspx

........................................
Chernobyl disaster..a must see pictorial

http://www.kiddofspeed.com/default.htm

 

[Deepseadata]

I'm scared even more now because it's seems like my 2tb's weren't the best idea.

I only have a max of 7 users on the network at a time and a total of about 14 user and email accounts configured.

Is it possible to just mirror the OS 146GB and mirror the Data 2tb?

I don't think I'm going to make an image for the OS and move it. I'm going to take the Exchange and User Files to another drive and point the applications at the new location of the databases.

That was the original plan anyway. That's why I came here. I needed a buddy to tell me if that was what he'd do... and I got a few!

Thanks guys. I'm listening to everything you guys say and learning a lot!

 

[goombawaho]

I'm scared even more now because it's seems like my 2tb's weren't the best idea.

Why??

technome - I don't believe that this issue is something that happens very often in the real world. This is more like theoretical failure mode analysis. It seems like your trying to scare the poster from proceeding with a very unlikely situation.

From my experience, I wouldn't worry nor would I deviate from the plan. Nothing can happen to your original data if you are only creating an image (disk to image). It cannot possibly get screwed up because you can't create an image on the same drive as your source (source and destination must be different). It's only during a disk to disk or partittion to partition that you can get them backwards.

If you first get a backup, image the drives, replace the drives, put the image on the new drives, there is really no way to lose any data. At the very worst (lighting strike??) you might have to restore from your data backup.

To the OP - if this is getting to be too much for you, bring in a professional that will take responsibility or at least get an "upgrade scenario analysis" (maybe for free) from them.

Don't know what else to say.

 

[technome]

Deepseadata...
Since you presently have 146 Gig drives, unless your hiring 100 new employees, do you really need 4 TB drives; just say you manage to come even close to filling the array to capacity, how would you back it up? If you use 4TB drives, rebuilds will take forever.
How about 4 SAS or sata in the >300 range, plus a hotspare, in raid 5. If you go raid 1, "reads" are only obtained from one member of a raid 1 only (with LSI based contollers), "writes" have a penalty, once the controller cache is flooded, as in raid 1, writes are committed to one drive, then written to the other member. Raid 1 is safe, and fine for a lightly used system, say for 20 users, where few reports/ searches are done, with a small number of writes at any given moment...basically, if you do not expect your disk subsystem to be hammered on a continuous basis it is fine. With the raid 5, reads will be very fast, the "write" penalty of raid 5 will be compensated by the controller, unless the disk system is hammered, reads remain high. Disk systems generally service approx 80% reads, 20% writes.




........................................
Chernobyl disaster..a must see pictorial

http://www.kiddofspeed.com/default.htm

 

[Deepseadata]

Okay.

After doing some more reading I think I'm going to mirror both my 146gb OS drive and my 2tb data drive for a total of 4 drives.

We don't have a lot of users but we do have a lot of personal data like shared photos, video and music.

I'll put just the OS and apps on the mirrored C: and put the Exchange database and personal stuff on the Mirrored D:

Wish me luck!

P.S. - What imaging software do you guys like?