changing 18 Gb to 36 Gb disks for FC10

[Kirill Cherkashin]

Hello,

we have next configuration:
HP-UX11.11,L-3000 and FC10 with 6x18 Gb disks with mirroring using exlusively for Oracle datafiles.
This weekend we're going to upgrade our 18 Gb disks to 36 ones. Which best way to accomplish this?
It seems to me that he most stupid and more simple is next method.
Stop oracle,backup oracle files to another location,put new disks in FC10,change volume group parameters and restore data on new disks. That's it.
It's not a big problem for us stop database because we also going to add more RAM and in any case we have to reboot our box. However,I'd like to know just for future (very soon we also going to upgrade another machine) how is possible to do it in more right way with minimum downtime.

For instance,could I put new disks in FC10 and afterwards add them to logical group as second mirror and after syncronisation remove one by one old disks? or even more could I replace old small disks by bigger ones on fly?

Thanks in advance,

Kirill

[harry d brown jr · Accepted solution]

Kirill,

The "best" way is ==> the way that MINIMIZES unexpected downtime, and you already have those procedures. Like you said: shutdown database, back it up (BUT I'd DO IT TWICE and check the BACKUP), then swapdisks, and rebuild your VG.

live free or die
harry

[Ceesjan van Hattum]

What about this (assuming
you have OnlineJFS):
([D]=data; [E]=emty)

unmirror 2x18 to 18[D]+18[D]
replace 1 disk: 36[E]+18[D]
mirror/synchronize: 18[D]+36[D]
unmirror disks
replace 2nd disk: 36[D]+36[E]
mirror/synchronize: 36[D]+36[D]

Now, both new disks uses old vg/lvol, but now with OnlineJFS you should be able increase PE,FS,VG,LV and so on....

This methode has minimal downtime, but... always make sure you have doublechecked your backups ... :)

[Sandip Ghosh]

Everything depends on how you had created the Volume Group Initially. I mean to say , what was the max PE value for the volume Group? If it is big enough to hold your 36 GB disk you can do it on fly means removing the mirror disk and again resync it.

You can check the Max PE of your VG by vgdisplay

Sandip

[Ceesjan van Hattum]

appended:

In the latest versions of Oracle, it is possible to freeze the DB on one of mirrored disks. For maximum availability: use 3 disks.
Freeze the database on 1 of disks, the other 2 will keep running mirrored.
Replace the third and re-mirror. Meanwhile the DB can be updated, so after syncing the disk, also sync the DB (!).
Repeate this procedure for all disks... maybe a lot of work, but low risks..
Succes to find your strategy!

[Gerardo Mora]

The problem using mirror will be to use the full 36Gb disk size, because the Max PE value for you volume group will be to 18Gb disks (like 4350 Max PE), you can check it with vgdisplay -v vgXX, I think you will need to re-create the volume group.

[Kirill Cherkashin]

thanks to everyone.
as usually more simple way is the best one.
;-) so,harry's answer is a winner.
unfortunetely,ceesjan suggestion doesn't work for us yet because installing OnlineJFS and mounting with its options is minor goal of upgrade.;-)
anyway, thanks again!