TMS
asked on
RAID Volume expansion
I have two clients with HP Proliant servers that come with an embedded software RAID controller (B110i). The controller is entry level so can only be configured for one array and two logical volumes. The current config is two mirrored 250GB HDD with two similar sized volumes.
I added two new 1T HDDs and then realised the controllers limitations.
So my question is around the easiest way to use the new HDDs.
Option 1 is to use my backup app BackupAssist which can perform a bare metal restore to different hardware. This is straight forward but I'm not sure how the volumes would be expanded. I'm thinking physical data location on the disk here. The server runs SBS2011 so I don't know if this would create unnecessary disk activity. Downside is I loose the 250GB HDD.
Option 2. This is my theory here. Copy off the data from the second volume, which is programs and Exchange mail database and then deleted the second volume from the raid controller and expand the first colume to use the newly freed space. Then add the 1T HDD to the controller as a second mirrored volume. Restart the OS in safe mode and expand the first volume so to include what was the original second volume and then initialize the new 1T second volume and give it the same drive letter and the original second volume. Then copy back the program and Exchange data and restart the server in normal mode.
So the question is that would both these options work. Option 2 is my preferred end state.
Thanks
I added two new 1T HDDs and then realised the controllers limitations.
So my question is around the easiest way to use the new HDDs.
Option 1 is to use my backup app BackupAssist which can perform a bare metal restore to different hardware. This is straight forward but I'm not sure how the volumes would be expanded. I'm thinking physical data location on the disk here. The server runs SBS2011 so I don't know if this would create unnecessary disk activity. Downside is I loose the 250GB HDD.
Option 2. This is my theory here. Copy off the data from the second volume, which is programs and Exchange mail database and then deleted the second volume from the raid controller and expand the first colume to use the newly freed space. Then add the 1T HDD to the controller as a second mirrored volume. Restart the OS in safe mode and expand the first volume so to include what was the original second volume and then initialize the new 1T second volume and give it the same drive letter and the original second volume. Then copy back the program and Exchange data and restart the server in normal mode.
So the question is that would both these options work. Option 2 is my preferred end state.
Thanks
ASKER
Thanks. The B110 does allow expansion according to the User Guide, but its not a great document. And in my ACU the wizard does offer the option to add the new disks to the Array, though I'm not clear what the result would be.
I do need to extend the C volume as SBS is hungry and I should of added more initially. I was thinking about adding a P model controller but it would only solve half the problem.
I guess I could use the bare metal restore option and restore the system to a newly created array volume using all of the original HDDs and then restore the data to the new HDD on the second volume. This way i could expand the volume in Windows and also could leave the Exchange data on the c drive if this improves performance. I can't reinstall Exchange as it will screw up the SBS wizards and scripts.
Taking your line I could use a P type controller and put the data on the new disks too if it isn't better controller. If I am to change then I'd prefer a hardware controller.
Another question is if I did use a P series, could I just plug in the current disks and would they be recognised? If so then you suggestion would be the easiest as I could configure accordingly.
I do need to extend the C volume as SBS is hungry and I should of added more initially. I was thinking about adding a P model controller but it would only solve half the problem.
I guess I could use the bare metal restore option and restore the system to a newly created array volume using all of the original HDDs and then restore the data to the new HDD on the second volume. This way i could expand the volume in Windows and also could leave the Exchange data on the c drive if this improves performance. I can't reinstall Exchange as it will screw up the SBS wizards and scripts.
Taking your line I could use a P type controller and put the data on the new disks too if it isn't better controller. If I am to change then I'd prefer a hardware controller.
Another question is if I did use a P series, could I just plug in the current disks and would they be recognised? If so then you suggestion would be the easiest as I could configure accordingly.
If you want to minimize downtime, why not add 2x1TB drives as it's own RAID-1. Move the Exchange database, any SQL databases, and user/shared data folders to the new volume.
This will immediately empty out your system volume. Does not require a backup/restore cycle, and all the downtime. You should still do a full backup per your regular schedule.
You can continue to use the existing system volume, and expand in the future if you need more space.
This method should not require a reboot. Just stopping services, moving files, changing config to point to new file location, re-start services. Downtime for users/services, but minimal when compared to a bare-metal restore.
This will immediately empty out your system volume. Does not require a backup/restore cycle, and all the downtime. You should still do a full backup per your regular schedule.
You can continue to use the existing system volume, and expand in the future if you need more space.
This method should not require a reboot. Just stopping services, moving files, changing config to point to new file location, re-start services. Downtime for users/services, but minimal when compared to a bare-metal restore.
ASKER CERTIFIED SOLUTION
membership
This solution is only available to members.
To access this solution, you must be a member of Experts Exchange.
you should be able to try modified versions of your second solution.
if you can only create one array, 2 ideas come to mind
A
- remove one of he drives
- add a 1TB disk
- give it enough time to resilver
- replace the remaining drive with the second 1TB
- resilver
- expand
B
- move one of the drives from the raid controller to a regular port (if not feasible, revert to a block-level backup and proceed in a silmilar way as below)
- create the arrays and volumes as required
- dd the existing partitions to the new ones using whatever live cd you are comfortable with (best is to use gparted that features partition copy, resize, ntfs resize...)
- expand the filesystems to fill the new partitions (gparted is still your friend)
- reboot
the first one should be feasible without any downtime or reboot but you'll get lame performance for a long time. the second is easier, probably more secure, most likely much faster but you'll have no service during that time
---
btw, using hardware raid for mirrors or stripes is pretty useless so if you can plug the disks directly on regular ports, you can setup software raid in windows and do similar manipulations with less hassle, less risks, and leave your working system alone without loosing the 250Gb disks
if you can only create one array, 2 ideas come to mind
A
- remove one of he drives
- add a 1TB disk
- give it enough time to resilver
- replace the remaining drive with the second 1TB
- resilver
- expand
B
- move one of the drives from the raid controller to a regular port (if not feasible, revert to a block-level backup and proceed in a silmilar way as below)
- create the arrays and volumes as required
- dd the existing partitions to the new ones using whatever live cd you are comfortable with (best is to use gparted that features partition copy, resize, ntfs resize...)
- expand the filesystems to fill the new partitions (gparted is still your friend)
- reboot
the first one should be feasible without any downtime or reboot but you'll get lame performance for a long time. the second is easier, probably more secure, most likely much faster but you'll have no service during that time
---
btw, using hardware raid for mirrors or stripes is pretty useless so if you can plug the disks directly on regular ports, you can setup software raid in windows and do similar manipulations with less hassle, less risks, and leave your working system alone without loosing the 250Gb disks
The easist option is to backup the existing configuration in Windows. Then rebuild thenRAID with new drives and restore from backup with autoresize during restore so that it would take new space automatically.
ASKER
Guys,
Thanks for all the comments/suggestions. I spent today looking at this a bit more. This was my first server build and what I realised is that I configured the volumes on the RAID controller. If I hadn't then I would have just been able to add a second array. I think I might have realised this as I looked at a similar set up on a later server install and on that one I had used one array for the two disks and then created partitions in the OS.
Andyalder, you are correct, the extend option does require a battery, so that option is out. Thanks for noticing.
Skullnobrains, I'm curious to understand why you think the OS level mirror is preferable to a controller mirror (which for the B110 is a also software RAID)?
I think I'm going to experiment as far as I can on my test server as my client trades 6 days a week so I have a reduced window of opportunity.
I can easily replicate a recovery from a backup to different hardware which was my first option. This gives me an additional thought which is to copy off the E:\ volume and then do a backup - then delete the arrays and re-create just one array and restore to that new single array (assuming the backup doesn't know the raid exists - I'll find out). This, I'm sure you are saying gets me back to where I started.. However, once restored I could delete the E:\ volume and then extend the C:\ volume in the OS which I'm guessing the OS will be able to do. If I then create a second array with the new disks and name it E:\, the same as before, and copy the files back to this new volume then this would give me the 250 + the 1T with the same config.
I'll also try the gparted and hope it is friendly...! I've not used it before so would be interested to try. I would be concerned about any possible prolonged performance degradation given the limited time.
To limit disruption then Andyalder's and Aleghart's option of just putting in a new controller works and then I could move around as suggested.
Whilst writing I have also just thought that maybe I should be less worried about the 250 drives and just restore to the 1T config. If it didn't work then I could just pop the old drives back in and re-plan. My initial plan was to move the NAS (legacy unsecured single disk) data on to the new drives with the Exchange data but I have also set up iSCSI on two new NAS drives so I may put the data on those (I want to enable DFS between the offices) and then just use the server disks for the OS SQL db's and Exchange.
I'm going to accept all the solutions as all have merits and clearly I need to be a bit more comfortable before approaching this.
Thanks again.
Thanks for all the comments/suggestions. I spent today looking at this a bit more. This was my first server build and what I realised is that I configured the volumes on the RAID controller. If I hadn't then I would have just been able to add a second array. I think I might have realised this as I looked at a similar set up on a later server install and on that one I had used one array for the two disks and then created partitions in the OS.
Andyalder, you are correct, the extend option does require a battery, so that option is out. Thanks for noticing.
Skullnobrains, I'm curious to understand why you think the OS level mirror is preferable to a controller mirror (which for the B110 is a also software RAID)?
I think I'm going to experiment as far as I can on my test server as my client trades 6 days a week so I have a reduced window of opportunity.
I can easily replicate a recovery from a backup to different hardware which was my first option. This gives me an additional thought which is to copy off the E:\ volume and then do a backup - then delete the arrays and re-create just one array and restore to that new single array (assuming the backup doesn't know the raid exists - I'll find out). This, I'm sure you are saying gets me back to where I started.. However, once restored I could delete the E:\ volume and then extend the C:\ volume in the OS which I'm guessing the OS will be able to do. If I then create a second array with the new disks and name it E:\, the same as before, and copy the files back to this new volume then this would give me the 250 + the 1T with the same config.
I'll also try the gparted and hope it is friendly...! I've not used it before so would be interested to try. I would be concerned about any possible prolonged performance degradation given the limited time.
To limit disruption then Andyalder's and Aleghart's option of just putting in a new controller works and then I could move around as suggested.
Whilst writing I have also just thought that maybe I should be less worried about the 250 drives and just restore to the 1T config. If it didn't work then I could just pop the old drives back in and re-plan. My initial plan was to move the NAS (legacy unsecured single disk) data on to the new drives with the Exchange data but I have also set up iSCSI on two new NAS drives so I may put the data on those (I want to enable DFS between the offices) and then just use the server disks for the OS SQL db's and Exchange.
I'm going to accept all the solutions as all have merits and clearly I need to be a bit more comfortable before approaching this.
Thanks again.
SOLUTION
membership
This solution is only available to members.
To access this solution, you must be a member of Experts Exchange.
SOLUTION
membership
This solution is only available to members.
To access this solution, you must be a member of Experts Exchange.
How about option 3, add a Smart Array P400 so you can have as many arrays and logical disks as you want? Under $100 on eBay and gives you the speed advantage of having Exchange data on one set of disks and the data on another set. N.B. I don't think the limit is one array, I believe you can have two RAID 1s with the current setup but a backup/restore would be needed without logical disk extension.