13 replies
by jdclyde
and then just restore to the new drives.
I would look at adding a few more drives to a second controller before I would just throw the old ones away.
If your running low on other resources (cpu, memoery) you may be better off with replacing the whole system.
by liame
More details.
What OS? What is the current disk setup? IDE or scsi? What flavour of scsi? What controller? What level of raid? What room is available in the machine? What end result do you need to achieve? Speed, redundancy, cost or a balance? Is it just data or system disks we are looking at here?
by 1ronman
by liame
5 sets of mirrored drives? Belt and braces huh?
If the drives are hot swappable you should be able to do this without anyone even knowing.
Firstly - get a good backup. Secondly - verify that backup. Third - see points one and 2.
Do it out of hours or at a quiet time if possible.
Lets call your sets 1-5 and the disks in each a and b. Pop out disk b on set 1 and replace with bigger disk. Rebuild mirror. Repeat for disk b of the other sets. Then back to the first set and disk a this time - replace all the a disks, rebuilding the mirrors as you go.
As these arent system drives you will be able to expand the volumes into the free disk space using disk manager.
by vinc_1969
by 1ronman
by liame
Well as you told me its data disks we are looking at not the disks containing the OS you can convert them to dynamic from to enable this functionality. If the mirror is software based rather than using the perc's they already are dynamic.
If it were the system disk(s) we were talking about here you would indeed 3rd party software such as partition magic.
As a tip I'd suggest popping a few of the new disks in a test machine when you get them and playing around with the various options and functionality in a safe environment until you are confident to take this over to a live server. There are few easier ways to lose data than messing with partions and getting it wrong.
Get a back up and verify it before starting.
by 1ronman
Thank you!
GB
by featherman
... that many RAID controllers will not accept the procedure Liam describes, as they will allow you to break and rebuild the mirror sets but will RETAIN the ORIGINAL volume size (ie: if the original volume /disk was 36GB and you break the set and replace one disk with a larger disk and rebuild the mirror, the resulting volume on the new disk will be the same as the OLD/ORIGINAL volume). This holds true for Adaptec, LSI and 3ware at least... Check with your documentation or RAID controller manufacturers' tech support to confirm...
If you perform the migration by simply replacing drives and re-creating volumes as previously suggested, and then utilize the Windows disk management MMC to grow the partition, you will have a sort of "hybred" RAID, where the actual RAID/mirror relationship is controlled by the RAID comntroller, and the volume size itself is controlled by the OS. NOT a situation which is pretty to support if anything goes amiss...
The best/most efficient way to achieve the results you are looking for is to perform a full VERIFIED backup of ALL of your data on a given array/arrays, replace disks A and B from any given RAID set, create the RAID relationship IN HARDWARE at the new volume size on the new drives, and restore the data to the newly created volume. Bear in mind that you WILL need to import the new volume into Windows via disk manager, and most likely delete the original volume (which Windows will most likely still show as existant but inaccessible).
The above procedure is as per standard/accepted practice from/of a major RAID controller manufacturer, for which I performed Technical Suppport in another lifetime....
Note as well, that there are sometimes powerful command line interfaces (CLI's)which might allow you to first break the RAID volume, re-create the relationship using one of your larger drives, break the relationship again, install a new second large drive, and grow the partition and re-create the RAID. I urge you to call your hardware vendor for support if you opt for this course of action... As I said before, the most efficient way to do this is Backup/restore...
Partition Magic may also be used, but I have had spotty success with RAID volumes and PM - it works much better where there is simply a physical hard disk with one or more partitions to be "played with"...
by advancedgeek
My experience with RAID reflects the cautions listed in this post. It's always cleaner to just do a complete replacement of the set as a whole. Plus, the disks will be defragmented when you restore your backup. ( Make two or three full backups and check their integrity before messing with the hard disks. )
I got a bit queasy reading LiamE's suggested procedure for this task. You ALWAYS have to count on Murphy's Laws to factor into any situation. Reconfiguring an Oracle server during normal business hours is an opportunity to make a HIGH VISIBILITY DISASTER happen. I always do this kind of thing after hours. If you break something at 8:00 p.m. you have twelve hours to fix it with no harm done. If you break something on Friday night you have two days to fix it. There have been times when I used most of that safety margin to fix some unexpected problem that cropped up during some 'routine' maintenance or 'simple' reconfiguration. After twenty years in the business I NEVER regretted the time that I spent after hours working on systems. I've had people laugh at me for my approach. Sure you can shoot from the hip and do things on a system while people are using it for business and 99.9 percent of the time you might get away with it. But some day taking risks during business hours will bite you in the buttocks. It's bad enough to break something. It's another order of magnitude of horror to have the business stop and have the managers looking for you just because you didn't want to work after 5:00 p.m. or work on Saturday or whatever.
by jtowner
by liame
All of Featherman's points are very true, nice to see you were paying attention.
I should really point out that the answer was specific to a PERC controller / dell servers setup as they come with the array manager software. This allows for the easy breaking, rebuilding and resizing necessary to make my method work.
If downtime isnt an issue I'd probably go with the backup /restore method myself but thats not always possible. In this case there is little chance of data loss as there are backups (none of us were joking about backups!) and the mirrors themselves, all of which have a full data set. In any case there is something satisfying about changing disks with zero downtime.
Running the virtual disk reconfiguaration wizard in array manager(right hand click on the appropriate array and choose 'reconfigure') allows you to add space to an array which you can then use diskmanager to 'grow' into safely without any of the 'hybrid' headaches Featherman pointed out (correctly) as possibilities. Indeed I have doubts if it would work at all without the array manager step.
Sorry for missing out an important point and step! I was just checking that everyone was paying attention.
This discussion is closed to new replies.
Trending Storage posts
Sort by Trending