MCSG and raw device file

[Mary Ann Lipa]

Hi,
we have MCSG A.11.15
we are planning to configure a package with a sybase/raw device file.. is it supported?
any special instructions?

thanks!
ann

[G. Vrijhoeven]

Hi Ann,

I do not think it works any different. Just do not create a filesystem on the cluster awaire volume groups logical volume.

Regards,

Gideon

[Dietmar Konermann]

Ann,

No problem to use raw devices/rlvols of clustered volume groups from a support perspective.

However, take care that your customer_defined_halt_commands() stop really *ALL* processes that keep the devices busy. The fuser(1M) loop in umount_fs() only recognizes mounted file systems... so a busy raw device would prevent the package from halting successfully, since SLVM is unable to deactivate the associated volume group.

Best regards...
Dietmar.

[Shaikh Imran]

hi,
You can very well use raw devices.
The recovery process will not be lengthy as fsck is not used for raw device files.
In fstab the entry will be like:
/dev/vg01/lvoldb raw_tables ignore ignore 0 0
OR
/dev/vg01/lvoldb raw_free ignore ignore 0 0

Regards,

[Mary Ann Lipa]

Thanks all!
what about the package control file?
there's an FS() parameter on the cntl...
do i disregard or remove that entry?

thanks again!
ann

[G. Vrijhoeven]

Hi,

Since it is not a filesystem delete the entry. You do not want this mounted.

Mind Dietmars comment on the halt commands. Make sure all processes are stopped.

Regards,

Gideon

[Shaikh Imran]

Hi,
I think the cluster aware volume group option will override the filesystem option.
so you can eliminate the
LV() FS() FS_MOUNT_OPT()


Regards,

[Kent Ostby]

HP also sells an enterprise toolkit that has scripts for Database setups.

http://www.software.hp.com/portal/swdepot/displayProductInfo.do?productNumber=T1909BA

[Steven E. Protter]

Your raw file is supported.

All you need to do is make sure ownerhip is consistent between the machines.

My raw file database is owned by the user sag.

So long as the user exists on both systems(it would be best if it had the same user number) everything will work well.

SG shares the disk. When a package fails over it has access to the exact same disk(raw as in uncooked) when it fails over.

SEP