New disks are soon arriving and i want to move a InformixServer 9.20HC3 server from old disk to new one.
I have got about 70 giga of data and I don’t want to download data with unload or HPL.

So i have tried the unix command dd to copy data from an old raw device to a new and it works with data ( I have not tried it yet for rootdbs ). The server start with the new chunk and I can do sql queries.

But I must keep the old chunk name ( chunk_6_11_1 ) which is outdated ( it’s the old name of the raw device ) an it’s going to be difficult to maintain.


Does anyone know a better way to move a dbspace from a chunk to another ?

Thanks

In IDS 9.40 you can do a redirected restore which means
that you could rename your chunks during a restore.
The old and new chunks names have to be specified in
a 'chunk mapping file'.

In versions < 9.40, you will not be able to rename your chunks.
You could use the dd-approach or your logical volume manager
to move the data from the old disks to the new ones.
Define the old chunk names as symbolic links to the new chunks.
You might also take a Level0-Backup of your data before doing
this.

Instead of using 'dd', you can also restore your Level-0-Archive (after setting up the symbolic link to the new chunks). However a 'dd' might
be faster.

I have also seen customers, editing the rootdbs reserved pages thru
an hex editor and fixing the chunk pathnames. This might work as
long as the new chunk pathnames are smaller as the old ones. But
I would never recommend to do it that way. The risk of damage is
definitely too high.

backup and restore with symbolic link(old name new target)
or unload-load (not recommend)

Thanks you very much for all your advices,

- Since I have no problem with 9.20HC3 version and I have got 4gl and esqlc programs I don’t want to upgrade my version of Informix Server.
Am I wrong?

- As you said, hex editing the reserved pages is not for me and I am afraid by consequences.

- Using the Level-0-Archive is a good idea.
Perhaps the benefit of using dd instead of Level0Archive is also that you can do the move in several times , moving some chunks and leaving others .

About the time with a big block size ( bs=2457600 ) it took me about 4 minutes for a 2 gigabytes chunk but I have no idea about the total restore time with a Level0Archive.


Also I want to take advantage of the move of chunks to get rid of the Informix Mirroring (which have cause some trouble ) and to work with the new disk with HpUx lvm mirroring.
I think about just starting the server without Informix mirroring, stopping it and move data .

If you use the dd approach make sure you CALCULATE THE OFFSET IN YOUR COUNTS!!!!!!!

I recently lost a few pages of data because for some reason some idiot made the offset of a few raw partitions 1 instead of 0. end result was i lost the last page of data on those chunks.

cool, huh?

and convienantly I lost one more than ontape allows before aborting.