Can Ignite-UX restore system running MC/SG?

[Gino Castoldi_2]

Hi,

Mgmt Svrs: L2000 (2 nodes) HPUX 11.0
MC/SG 11.09 Oracle 8.1.7.4 OVO 6.14 NNM 6.1
MirrorDisk/UX AutoRAID 12h disk array.

VG=vg00 for the two mirrored system disks.
VG=vg01 for the shared disks/vg's.

We are using Ignite (B.4.2.110) to make backups
of the entire system with these commands:
make_tape_recovery -x inc_entire-vg00 -x inc_entire=vg01

Will we be able to restore/recover the entire
system if there are any problems?

10 points to any good answers.
TIA, Gino

[Mark Grant · Accepted solution]

It is actually possible to use Ignite to backup a cluster but I would start to worry that you might not get vg00 and vg01 on you backup medium. I think restoring of this would also be much easier if the cluster is down when you take your Ignite backups too. I think you might have some problems with newly created vg01's disks not being cluster aware if you are restoring completely cold.

We have our ignite set up so that when we do a disaster recovery, we don't activate service guard at all. We would rather just get the volume groups activated, the filesystems mounted and the applications running. In a DR situation we feel it will be quicker to get up and running this way and we can worry about getting the cluster working again afterwards.

Not saying it's good but it IS simple :)

[Geoff Wild]

Of course you can!

Though I question why you make_tape_recovery vg01? Is this a cluster aware vg and is Oracle installed there and running when you make_tape_recovery? I don't know for sure that even if you put the DB in hot backup mode, if you will get everything on the tape correctly...do you also have another backup solution in place for backing up the DB? If yes, then no need to make_tape_recovery vg01....

Rgds...Geoff

[Gino Castoldi_2]

Hi,

When I run Ignite-UX I first stop all applications: shutdown Oracle listener and the db, stop all OVO and NNM processes and
gui's. I put the cluster into maintenance mode first so the shared disks are mounted.

I tried to do this in single-user-mode once before but I had trouble mounting the shared disks without the packages (OpC/NNM)running.

This is a production system so I didn't want to spend too much time with the application
offline trying to back it up. There must be a way but I am not sure how to do it.

10 points to any good answers.
TIA, Gino

[The Real MD]

None that I can see. its good that the databases, the listener and apps are shutdown first. It might take a while to generate the tape depending on the size of vg01 it also may well need to span tapes

I would certainly test the tape/s

Regards

Martin.

[Bill Hassell]

From a purely performance point of view, Ignite/UX will not run nearly as fast as fbackup. fbackup starts as many as 6 separate reader processes to keep the tape busy. make_tape_recovery just uses tar or optionally cpio, neither of which is a high performance backup tool. make_tape_recovery (MTR) is designed as a disaster recovery mechanism and is terrible when it comes to getting a single file or directory. Terrible in performance (will always read the entire tape) and terrible in that you must use special mt fsf commands and the right device file to get to the file backup portion of the tape.

I would make a disaster recovery tape using MTR once a week or two, and use fbackup for files on both VG00 and VG01. frecover has the ability to jump forward at 100x normal speed to recover a single file.

[Bernhard Mueller]

Gino,

I think Bill hit the most important issue, i.e. even though you can include vg01 in the recovery image, the time needed to restore would be inacceptable.

It is much better to have only vg00 in the recovery image and have a more performant backup of vg01.

You also need to consider that after recovery you still need to change the vg01 to be cluster aware again, before you can start the package.

But since you have a cluster, I wonder what you are trying to achieve at all? If you loose all HW where do you do a restore? Either you require a campus cluster with nodes and mirrored storage at two different sites, or you will need quite some time to get new HW to get a whole infrastructure up again anyway.

Regards,
Bernhard

[Gino Castoldi_2]

Hi,

Bernhard:
>You also need to consider that after recovery you still need to change the vg01 to be cluster aware again, before you can start the package.

How do you change vg01 to become cluster aware?

>But since you have a cluster, I wonder what you are trying to achieve at all? If you loose all HW where do you do a restore? Either you require a campus cluster with nodes and mirrored storage at two different sites, or you will need quite some time to get new HW to get a whole infrastructure up again anyway.

I agree, we have only one system so if we did lose all of the H/W then we have a serious outage (time-wise) because it's basically a SPOF. We have no alternative but to use this one single system for monitoring.

10 points to any good answers.
TIA, Gino

[Bernhard Mueller]

Gino,

vgchange -c y vg01

Regards,
Bernhard