Can you swap out bootable disks with Disk Admin striping?

[Hanson]

Okay,the header is a tad confusing. Here's the scoop:

I want to have a three disk server -- two will be used for striping w/o parity. The other disk would hold the OS and a daily backup file. The OS disk would hold no other programs while the striped pair would hold only data.

The question is: if the OS disk were to quit, could I simply pop in another disk and be good to go or would the striped pair fail to work now that the orignal OS disk information is gone?

The fall back option is to have an external disk hold the backup. This server is strictly for data only and is necessarily low maintenance since no MIS staff will be on-site for tape backups and etc.

Hanson

 

[Jpoandl]

Hanson,

I guess this server is very low priority. I am also assuming there is no hardware level RAID controllers on this system.

You have the following disks:

Disk 0
Disk 1
Disk 2

Disk 0 will be used for the OS. (No mirroring or any level of data fault tolerance.)

Disk 1 and Disk 2 will be used to store data. These drives will be striped together by the OS. (two 4gb drives, for example, making an 8GB volume) Also no fault tolerance.

If Disk 0 fails, you will not be able to boot the system. Therefore, you would either have to rebuild NT server or restore the OS from a tape backup. If you rebuild the server from a tape backup, you will have no problems. (other then server down time.)

However, if you have to rebuild the server from scratch, I think you will run into 100% data loss. I've never tried this because in my environment, data is safegaurded....all disks are fault tolerant. Anyway, I think you will lose all data based on two things:

One: If the server was built and LOCAL users/groups were created, this information would be lost during the server rebuild. All security information based on these users/groups will be lost on the stripe set volume.

Two: The stripe set is configured through NT Disk Administrator. The information with regards to the stripe set is held within the registry. Once the original registry is lost, NT will not know about the stripe set. After failure and the server OS rebuild, this registry information will be lost. Therefore, you will not be able to access your data on the stripe set. (You may be able to work around this problem is you backup the registry and have a good recent RDISK available....but still this restore will be difficult.)


If I were you, I would add one more drive to the system. Make an OS mirror for the OS drive. In this way, your system will be much more stable.

I would be interested in hearing other view points concerning this matter. As I don't run into this type of low level method of building a server.

Joseph L. Poandl
MCSE 2000

 

[Hanson]

Yes, the server is low priority. It is a backup server for our design department, and won't handle anything other than file storage. And we will be using IDE rather than SCSI drives. This means that adding one more drive would bump the CD-ROM off of the server. There is no tape backup.

Considering this server will only be accessed by 5 to 7 users (and each user would access files a few times a day), do I even need disk striping? I'm starting to think that the DTR of a single 5400 rpm disk will outstrip our network throughput anyway, especially considering it will be rare that more than one user will be accessing the server simultaneously.

Are there hitches or problems associated with striping without parity that I should be wary about?

If I keep the striping, the other configuration I am considering is this:

Partition 1 on IDE 0 holds the boot and OS. Partition 1 on IDE 1 holds the OS mirror. Partition 2 on IDE 0 and IDE 1 stripe data. IDE 3 is for backup.

Any thoughts?

 

[Jpoandl]

I have used striping without parity once...and well, I did not like it. There is no fault tolerance. Performance was good. But in your situation, you probalby wouldn't benefit from the gain.

The problem I saw with striping, is when the server is hard shut down (without shutting down properly...) there is a chance of losing the data. I've noticed that after the server is rebooted after a hard shut down, the server will have to restripe or run Check Disk. The volume I created consisted of a 200GB partition...(I wouldn't ever reccomend doing this again..). Anyway, it would take over 2 hours to check the disk out before allowing boot.

How big are your drives? How much space do you need for the end users?

Maybe something like this would work?

9GB - 9GB - 9GB

A) Create an OS partition = 4 GB; OS mirror this between disk 0 and disk 1

B) Create a data store partition = 4GB; Mirror this between disk 1 and disk 2

C) Create another data store2/backup partition = 4GB's; Mirror this between disk 0 and disk 2

Tell your half you users to save data on B and the other half to save data on C.

Maybe this would work...of course you do lose a lot of disk space to fault tolerance... In this configuration, if one disk fails, make sure you replace it...because it will effect 2 OS mirrors. Still I think this will provide a good measure of fault tolerance. I think you should really consider fault tolerance in your planning...especially if you are not performing backups.

Hope this helps..



Joseph L. Poandl
MCSE 2000

 

[Hanson]

Well the idea would be to perform daily backups internally onto disk rather than tape because there won't be any MIS personnel on site to perform tape backups. We would also perform an off-site backup over the network once a week.

I am going to skip striping. However, if I mirror the data store, would there be a noticable performance hit as far as saving files is concerned?

 

[Jpoandl]

No, there would be less performance with mirroring...but for less than 10 users, you should be fine. What kind of hardware...I would thing that you should have at least a Pentium Pro 200 Mhz CPU for this. Joseph L. Poandl
MCSE 2000

 

[Hanson]

I'll be using a PIII 500 with 256 MB RAM.

But now I wonder... if I'm going to mirror the OS and the data, why even go with three drives? I can get by with two. And there's really no point in an internal backup, since all the data will be mirrored.

BTW, if the boot drive goes down, how easy is it to boot from the mirrored drive?

 

[Jpoandl]

Yeah, you might be better off with just two disks. Your machine will have no performance problems. That machine will be just fine.

You should also take other measures to protect you data. Backups are the best policy...but you should also create an Emergency REPAIR DISK after your mirror set up is complete. To do this just run RDISK /S from the RUN menu.
Make sure that you have a formatted floppy disk in the A: drive.

You should also create an NT boot disk. Here is how you do that:

http://support.microsoft.com/support/kb/articles/Q101/6/68.asp

This boot disk may be needed if your primary boot disk FAILS. You may have to boot the system using the floppy boot disk. (You may have to edit the boot.ini file to point to the other good OS copy disk. This is very easy..and if it ever happens you can always come back to this site to get the information you need to re-point the boot.ini)


The following is a link to MS all about common problems/repairs when using Disk Administrator:

http://www.microsoft.com/technet/tr...chNet/prodtechnol/winntas/support/ftshoot.asp

It is pretty easy to recover a system from broken mirrors. You should check Disk Administrator every so often for mirror problems. All mirrors should display "healthy" as the status. Joseph L. Poandl
MCSE 2000

 

[Hanson]

Thanks so much for all of your input. But I have run into yet another NT problem.

When I install NT, I run into the 1024 cylinder limit for the boot area. So I went into the BIOS and used LBA. Instead of 500MB, I had plenty enough room to install NT on a 2 GB partition. So far so good.

However, when I go into disk admin, both disks are a little over 8 GB. In reality, they're 20 GB disks. This is confirmed by the BIOS, but I don't know how to make NT recognize the extra space.

Is there a workaround for this or am I in a bind?

Hanson

 

[Jpoandl]

Sounds like a BIOS problem. When you go into FDISK do you see less then 20GB's per hard drive?

This is a very common problem when using IDE drives. I performed a quick search in the PC area of Tek-Tips. There are many people with the same type of problem.

I think you were smart by selecting LBA mode in bios. This should be correct. I have heard in the past that you may have to update your BIOS in order for it to "see" large volumes.

With my PC's at home, when I boot the machine, I can look at the information diplayed on screen during boot. It tells me that it detected the hard drives...and I think it tells me the size. Can you see the size of the hard drives during boot?

Check this link out: thread751-51214 Joseph L. Poandl
MCSE 2000

 

[Mubashir]

Now that you've overcome most of your problems, Hanson, you can make NT see the entire of your drives by installing service pack 4 or above.

Regarding FT, you've made the right choice keeping in mind all the circumstances. However, I wish to add a new angle to this discussion. Since you're using the software solution for FT, what if both of your hard disks go down at the same time. Either there's some hardware failure, virus, or any unknown factor which renders both HDs useless. What would you do for the data that has accumulated since the last backup? Thats about 5-6 days' data. Though, it sounds unlikely but it's possible, eh?

My idea of FT is that you should get the data 'out-of-the-system', either on tape, network drive, CD, or any other medium. The idea is the same as off-site storage of critical data. FT in the same system, IMO, has always some degree of risk attached to it.

For this reason, I would consider a rather less-explored feature of NT called 'directory replication'. But for this you would need one more NT server or workstation. One NT server configured as export server sends data updates to multiple servers/wks configured as import servers. In this way you can have up-to-the-last-minute updated data, should the export server go down at peak time.

I have used directory replication with 500+ MB of data on a LAN without a hitch. It's a nice, smooth way of keeping your data backed-up with virtually no server downtime. All that's needed then is to redirect the user links towards the new server. A lot depends on the kind and amount of data you have. Think about this possiblity, prepare atleast one server for data updates and you might like it someday!

Mubashir
muhammad.mubashir@sbp.org.pk

No one's worth your tears. The one who is, won't make you cry.

 

[Hanson]

Thank you for the input, Mubashir. We were debating on whether or not we'd need a third HD for backup, but your comments confirm that it is a necessity. In addition to virus contamination, we also ran into a situation where a user deleted all files by accident and needed them back. So it is still a good thing to have a backup.

We are going to use a tape backup once a week in the case of a catastrophic problem, such as server theft or fire.

As for directory replication, I will look into it as I have a workstation that would be perfect for it. The only hitch is that our network is extremely slow and will be until we can replace our hubs with switches. This is why I can't perform backups over the network.

And yes, the service pack fixed all the disk space recognition problems.

Again, thanks for all the help. Couldn't have done it without you.