Disk replacement/upgrade for all SVM mirrors
I am currently running Solaris 10 with two mirrored boot drives using SVM. The two drives used are one of the two internal drives and an external SCSI drive. The second internal drive is used entirely for /export/home. The contents of this disk are being moved temporarily to a NAS device thus leaving this drive empty.
Our current plan is to remove the non-boot disk drive and replace it with 147 GB disk drive and add this disk as a mirror using slice sizes identical to the current mirrors including 2 database copies (slice 6 & 7).
Then we are planning to remove the original boot drive from the mirror (c0t0d0), replace it with another 147 GB drive and then add this new drive as a mirror.
We will then remove the single external disk drive from SVM, leaving the mirrors only on the two internal drives.
We will then use the available space on the two 147 GB drives in slice 5 (unused currently) for /export/home.
I need to get any advice concerning the procedure outlined above. The server we are working on is our jumpstart server, our NetBackup master and media server and our home area file server.
Our current plan is to remove the non-boot disk drive and replace it with 147 GB disk drive and add this disk as a mirror using slice sizes identical to the current mirrors including 2 database copies (slice 6 & 7).
Then we are planning to remove the original boot drive from the mirror (c0t0d0), replace it with another 147 GB drive and then add this new drive as a mirror.
We will then remove the single external disk drive from SVM, leaving the mirrors only on the two internal drives.
We will then use the available space on the two 147 GB drives in slice 5 (unused currently) for /export/home.
I need to get any advice concerning the procedure outlined above. The server we are working on is our jumpstart server, our NetBackup master and media server and our home area file server.
ASKER CERTIFIED SOLUTION
membership
This solution is only available to members.
To access this solution, you must be a member of Experts Exchange.
> If slice 5 is the only unused slice, you may be stuck.
> In order to mirror the disks, you have to create metadbs, which are normally on their own slice.
Not necessary. If there are no free slices, metadb may be created on any of existing slices (except swap and entire disk), but _before_ this slice was formatted with 'newfs'.
> In order to mirror the disks, you have to create metadbs, which are normally on their own slice.
Not necessary. If there are no free slices, metadb may be created on any of existing slices (except swap and entire disk), but _before_ this slice was formatted with 'newfs'.
oops, I meant _before_ this slice was added to metadevice :-)
>>Not necessary. If there are no free slices, metadb may be created on any of existing slices (except swap and entire disk), but _before_ this slice was formatted with 'newfs'.
Much like towing a semi-trailer with a pick-up truck, it might work, but it is very poor practice...
Much like towing a semi-trailer with a pick-up truck, it might work, but it is very poor practice...
If we have no other choices, there is not a case to mention good practices :-)
Since you are using a new HD, you can when you partion the new HD, you can create an extra small slice/partion for the metadb!
ASKER
In my notes I stated that for all mirrors the metadb's exist on slices 6 and 7, two copies per disk. Thus using slice 5 for data will not affect the SVM configuration.
Regarding the installboot step. I was under the impression that mirroring slice 0 (/) would take care of the boot block. Thus we would not have to run the installboot. Can you please comment on this.
Regarding the installboot step. I was under the impression that mirroring slice 0 (/) would take care of the boot block. Thus we would not have to run the installboot. Can you please comment on this.
If you break the mirror first then ufsdump/ufsrestore the system to
the new HD, you do need to run the installboot !
the new HD, you do need to run the installboot !
SOLUTION
membership
This solution is only available to members.
To access this solution, you must be a member of Experts Exchange.
ASKER
The system is an old Ultra 60, 2 CPU's and 2 GB memory.
I have already moved /export/home off the server and on to a NAS device. Thus I will just recover the area to the new slice 5 with find | cpio; this preserves permissions and links better than cp, mv or tar.
I will end up halting the server to the "ok" prompt in order to verify the disk aliases to update the boot-device eeprom setting
Thanks for all the assistance.
I have already moved /export/home off the server and on to a NAS device. Thus I will just recover the area to the new slice 5 with find | cpio; this preserves permissions and links better than cp, mv or tar.
I will end up halting the server to the "ok" prompt in order to verify the disk aliases to update the boot-device eeprom setting
Thanks for all the assistance.
In order to mirror the disks, you have to create metadbs, which are normally on their own slice.
You don't mention the model of the system. That would make a difference as to how I would do the upgrade. If it is one that allows the hot swap of the disks, I would set up the mirrors first, make sure that they resysnc, and then remove the old disk from the mirrors and install a new disk in its place.