Urgent: mirrored scsi disk upgrade problem.

[us2001irvine]

IBM NetFinity 86xx 8GB SCSI mirrored HDD almost runs out of disk space. I would like to upgrade them to two 16GB SCSI disks, and keep all the data from the current disk.]

Option 1: Server new install with application data backup restore pushing back in.

(This is not good choice right now, becasue I don't have good backup of that 8GB disk drive as well as I need to reinstall the Citrix Terminal Server and other application programs on that server.)

Option 2: SCSI disk imaging.

(This might be a good solution. Has anyone done this before, and how, and with what kind of disk imaging software being used, and will Symantec Ghost 2003 work? Any risk of doing this upgrade without backing up the whole server?)

Option 3: Can SCSI mirrored disk be volume-spanned with the rest SCSI HDDs on the server? I have 4 more bays on the server.

Option 4: ???

Thanks in advance for your kind help.

 

[DaveGoTech]

You didn't mention your O/S, but you should be able to add the larger drive and mirror it without using third party utility software.

 

[catorze]

Option 1 although not the most atractive from an implementation point of view is by far closest to the best long-term solution. Something to consider though is how much will your data grow over time in relatioin to the lifespan of the current server you are using. The best (easy) solution to implement would be to capture a complete backup of the server (then test restoring the data) and build a RAID configuration with HDDs larger than 16GB (IBM has SCSI HDDs from 18.2 GB - 146GB). With a RAID configuration, if your hardware platform supports more HDDs (you didn't indicate which server you are using), you can even define a Hot Spare to facilitate recovery for a HDD failure. RAID configurations also support Logical Drive migration to add space in the event your original guesstimate of how much space you need was wrong. Introducing software volume sets is the easiest means of defeating the redundancy you intended by mirroring the drives, and adds an additional layer of complexity (overhead) for the server to process - not a good idea for a Citrix server.

 

[us2001irvine]

1. It is Windows 2000 advanced server. (Sorry that I forget to include this.)

2. Good news: My client got 700MB free space after disk deframenting and reducing page files. Now free space is 800MB. And there is still rooms to clean up some users' big profiles of over 100MB. (Right now the server space is pretty ok.)

3. Long term solution: my client still wants to upgrade to 16GB HDD which are sitting on the shelf. For fresh re-build, it is fine except the Citrix server setup that I need figure out. For the disk capacity planning, 16GB will be good for a very long, long time.

4. For disk imaging, it is divided point of views about its feasibility. Anyone has tried this before and how? I appreciate your nice sharing, though I may not use this at this time.

5. For fresh rebuild, I will copy data to another disk. After fresh re-install of w2k server, Citrix server, and application programs, I will push the data back in the new-built server. Will this work?

Thanks both of you for your kind help.

 

[ibmtech65]

How are the drives mirrored? hardware or software?
If it is hardware you can pull one drive and then replace it with the new larger drive. It will rebuild the mirror using the new drive, but only the space of the smaller drive.
Once the mirror is rebuild pull the other smaller drive and put the other new drive in, it will rebuild the mirror.
Once the mirror is rebuild you can go into disk management and extend the partition to include the new free space created when you used the larger drives to rebuild on the smaller drive.

 

[catorze]

If you replace the HDD with a larger HDD, using hardware RAID, the RAID controller will assume the drive is the same size and the excess capacity will not be usable - you will NOT be able to partition the additional space.

Number 5, although the most tedious, is by far the best solution, since it will be a native installation of the critical components, and data transfer is pretty simple to do. This would be my recommendation for a long-term resolution.

 

[ibmtech65]

It can be done with a raid controller. You will see the free space in raidmanager you then create a new logical drive within raidmanger using the free space and then extend or create a new partition using that free space in disk manager.

 

[catorze]

May work as a last resort, but not a good solution to plan an upgrade. If you create a new logical drive, the OS will see it as a new physical disk. If you extend the partition (i.e. volume set), you have just undermined the purpose of having a RAID controller, and enabled the system to fail if a HDD dies - seen it happen more than once (the Volume set pukes, and you're screwed). If you create a new partition, you're better off doing it the right way the first time as I described above.

 

[us2001irvine]

Thanks for enormous input on the issue. It seems that we have three solutions: (BTY: it is hardware mirroring)

1. Fresh re-install and data restore: This might be the best selection, but it is time-consuming and error proning.

2. Hardware RAID disk swap with volume/partition reconfiguration: This is the most easy one and I will find out if it will work. (Any risk of doing this to the original disk. For the risk, personally I don't think so with my previous experience. But I really need to make 100% sure that I won't mess up the the original drive.)

3. Disk imaging: Cloning disk with larger HDD would be another quick and nice fix. Any comments and cautions on this? Also, without using disk imaging software, can I just copy-paste the whole HDD to another SCSI disk on the same server? It is just easy but I am not sure if it will work.

This project will be completed in about next 1-2 weeks. Right now we are waiting for the backplace for the last 3 SCSI disk bays. Thanks for your kind help.

 

[us2001irvine]

It is unavailable to update this problem issue at this moment because my client did not call back.

Thanks.