Our production server was only set up with one large partition.

I have set up a standby server and properly partitioned it. Now the boss wants the production environment's partition shrunk.

It is an HP DL380 G6

We have 4 hot swap drives in a RAID 5 array.

How best should I go about doing this. Seems like a bad idea to me.

Should I use windows or HP to do the partitioning?

What should I be aware of in a production environment? The idea is to put the site (Inetpub) on a separate partition instead of the C: drive.

How much downtime should I expect?

Is this a terrible idea?

Anything else I have missed?

EDIT/UPDATE: I presented the information I gathered and included the information given here by the users and we are NOT doing the partitioning! THANK YOU! FYI: His idea was to create two partitions so that if the OS failed the data would be safe.

Now the hardest part is going to choose a correct answer for this... You all gave such good information. Make a suggestion in the comments of this question and I will pick one of the answers.

  • Can I ask why you want to do this? I can understand that it is nice to partition your system files from your data files, but at this point it's already in production and working. Making the change will not improve anything so why does your boss want to do this at the risk of data loss and downtime? – ITGuy24 May 3 '10 at 20:13
  • The only reason I could see for this at this point is to make the database mirroring easier as I wont need to do a MOVE transact SQL statement for the mirrored DB being in a non similar path. But that is way easier than his plan. I am going to pop, I need help to demolish this plan. PS he does not know this would be a reason. This is my reason. I think he just likes the idea of DATA and OS separate. But why was this not done in the beginning then... – Campo May 3 '10 at 20:23

Um...not a good idea, in my opinion. It's a production server.

When would it be a good idea?

You would need backups. Tested that you can restore it from bare metal. I.e., you are playing with a partition. One mistake = lots of downtime.

You might be able to use a Linux bootdisk with gparted to edit the partition, but you're still taking quite a risk, esp. with the RAID controller.

Your best bet is to back up everything, wipe the drive and re-partition it, and restore data from the backup.

Better yet switch your standby server into your production server, then turn your production server into a standby and repartition it with the technique above.

Downtime-it could take an hour, or if something goes wrong, anywhere from half a day to a day or more. You won't know until you find out if something goes kaput.

Primarily I'd worry about migrating it off of being a production server before trying anything with it. You're kind of asking for trouble by playing with a live production server's partitions.

  • HEHE! thanks! I know.... Its so dumb. Problem is the stanby is the piece a garbage they replaced with the primary.... This has bad news all over it. Sigh. What do I say to the boss? I was planning on using "Shrink Partition" in the Computer Management MMC under Disk Management. Feel like I need to use the raid controller to do the partitioning... – Campo May 3 '10 at 19:11
  • Just checked the shrink partition option.... that's not going work so well. Not nearly enough free space. Plus I checked the HP array utility and as well there are no options to re-size. This seems like its going to be a big mess. Anyone else have any input? – Campo May 3 '10 at 20:03
  • 1
    Playing with partitions is never something to be taken lightly. Best way to do it is to wipe and reinstall from scratch; the only time you should play with a "live" system whose uptime is paramount in importance is if your door has a sign saying something like "neurology" or "cardiology" out front. Only "reliable" way for this is wipe, re-partition, restore from backup. This I'd budget a few days for just in case something went wrong, even then, not including the time to make backups and test them on a backup system. – Bart Silverstrim May 3 '10 at 23:28

Yeah, as Bart said, move the load over to the other server and then wipe and reload the first one in peace would be the simplest option. But I don't see much point either in having two partitions if they're both on the same RAID-5 array, fully extended - it's not like it will help performance or that inetpub is hard to find on the system volume.

I'd almost go the other route and redo the other server to only use one partition instead to solve both your path problem and the issue with tampering with a running production server? Or invest in a pair of extra drives for a separate operating system mirror array and wipe, reload them both ^^

If there's no space to shrink (love Server 2008+ volume management btw, it's finally beginning to do useful stuff like extend and shrink system volumes) you're out of online options. Use a defrag tool that will move files to the beginning of the disk - perhaps the built-in defragger would help you enough. Then re-run the shrink check. Have a well-tested backup and do it during a maintenance window.

  • I'd hasten to add that even with tools built into Windows (or any other tools), I'd never fully trust it paired with something with "production" in the label. :-) – Bart Silverstrim May 3 '10 at 23:29
  • 2 partitions on the same raid 5 will not give any benefit you say. HMM Then I must ask why he wants this so badly. I am confused. It never sounded good to me. The other server has only 2 mirrored drives. – Campo May 4 '10 at 0:41
  • 1
    Well he gets a logical benefit of separating out the data folder from C: - for whatever that's worth. It doesn't feel useful enough in this situation to me though as the performance won't really change. – Oskar Duveborn May 4 '10 at 6:39

Switch your standby server into your production server, then turn your production server into a standby and repartition it, while it might be junk it only has to work while you are rebuilding the old primary. If this is your bosses idea get him to approve what option he wants- use some planned downwtime and hope that the standby can be brought up if the server rebuild takes longer than expected or swap machines and take potentially degredated services while it's being built. This shouldn't be your call per se- present the options to the business and let them decide

  • +1 I was thinking that exactly on the way home lol! – Campo May 4 '10 at 0:40

First off, repartitioning is no big deal, with a number of readily available products to do the job, but there is always a possibility that something can go wrong. Murphy's Law ensures that the more you need this system up and running the higher the probability that it will all go pear shaped. You therefore absolutely must have backups that can be used from the cold metal state, as others have already said.

Despite the foregoing, you haven't given a real reason for such an operation, which can cause downtime lasting anything from half an hour to whatever time it takes to restore from the backup. Even when it all goes well the amount of downtime will depend on a number of factors, mainly the size of the partition, the amount of data on it and the level of fragmentation.

The old idea of separating the OS and data into separate partitions, at least on Windows machines, is only followed by a very small number of admins these days, as most of us learned a long time ago that there are numerous disadvantages and no real advantages. If they must be separated do so on separate drives, not partitions on the same drive. The use of RAID doesn't change this in any way.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.