make_recovery from old server to new

[Kieran Wareing]

Hi
I have a 6 year old two node cluster (2 x rp7400 PA-RISC) running HP-UX 11.00 with serviceguard and two old Oracle databases (Oracle 7 and 8). Everything runs on one node with the other idle.

If I had a new server (current hardware) for DR, would a make_recovery tape from the old server be able to restore onto the new one?

Would new hardware be able to run the old Oracle databases?
We are going to upgrade to Oracle 9 (we are always short of money!!) but probably not until the end of the year.

In any case, what server would you recommend for DR in this situation?

Thanks.
Kieran

[Torsten.]

I think your current server (what is the model?) won't run with the old and unsupported HP-UX version 11.00 - it will probably need 11.23 or 11.11.
So you cannot use ignite.

[gigiz]

what is the new server model's?
If it doesn't work with 11.00 you can do upgrade your old system from 11.00 to 11.11, and then migrate with make_tape_recovery (if it work) ...
:-)

[Kieran Wareing]

I have already investigated upgrading the OS on our existing servers and they haven't got enough root disc space so would require a lot of work and new root discs.

We have no new server yet because I didn't know if any would work with our existing old software.

So, since you say ignite will not work, I guess we would either have to source an old server that could run 11.00 or get a new one and install our software from scratch. But would new hardware run Oracle 7/8?

Thanks
Kieran

[gigiz]

For increase root file system you can do a make_tape_recovery and in RESTORE, adjust the file system size for following upgrade ....

[Kieran Wareing]

Unfortunately the existing system only has a 9Gb root disk. Here is PVDISP and VGDISP for it.

root> vgdisplay /dev/vg00
--- Volume groups ---
VG Name /dev/vg00
VG Write Access read/write
VG Status available
Max LV 255
Cur LV 11
Open LV 11
Max PV 16
Cur PV 2
Act PV 2
Max PE per PV 2500
VGDA 4
PE Size (Mbytes) 4
Total PE 4338
Alloc PE 4334
Free PE 4
Total PVG 2
Total Spare PVs 0
Total Spare PVs in use 0


Tue 24 Apr 07 [service1] /
root> pvdisplay /dev/dsk/c1t6d0
--- Physical volumes ---
PV Name /dev/dsk/c1t6d0
VG Name /dev/vg00
PV Status available
Allocatable yes
VGDA 2
Cur LV 11
PE Size (Mbytes) 4
Total PE 2169
Free PE 2
Allocated PE 2167
Stale PE 0
IO Timeout (Seconds) default
Autoswitch On

According to the 11i upgrade doc, /stand needs 100Mb free space and we only have 35Mb and /var needs 160Mb free and we only have 124Mb.

Kieran

[Patrick Wallek]

I know that Oracle 8i (8.1.7.4) will run under HP-UX 11.11. I don't know about Oracle 7 (that is getting pretty old). I would think that it *SHOULD* run, but the only way to find out is to test it.

[Bill Hassell · Accepted solution]

There are a number of ways to fixup the vg00 disk space issues. Naturally, your new server won't have 9Gb disks because they aren't manufactured any more (36Gb or 72Gb are typical). But the real question before you start is: what do you mean by "current hardware"? If you are talking about an Itanium Superdome, then the answer is no, you cannot restore the PA-RISC backup onto any Itanium server (anything with "rx" in the name). You also cannot restore onto really new models such as the rp8420 or rp7420 since these may need drivers installed on your old machine before you take the Ignite backup. Similarly, if you are using different I/O (SCSI vs. fibre) or different models of fibre cards, you may need to install the needed drivers before you take the bacakup.

As far as Oracle 7 or 8, you have to check with Oracle. 7 and 8 will never run on Itanium and support for 11.11 may be in question. For sure, 7 and 8 won't run on 11.23 or 11.31 OS versions.

[Kieran Wareing]

Thanks Bill.

So, forget make_recovery. Let's say we start with a PA-RISC server running 11.11 and let's assume for a moment that it will run Oracle 7 and 8.

How much memory does the OS need?

We would need 2 Gb memory for the databases and 2 73Gb disks. It would also need a DDS tape and a DLT tape for DataProtector 5.50.
All users will connect over the network with PC clients.

Can anyone recommend a server model?

Thanks
Kieran

[Mike Shilladay]

Hi Kieran,

Is it absolutely imperative that you have a new server? If you aquired a second user rp7400, (which will probably not cost much - most likely less than an annual DR policy and the asset is yours to keep) you will save yourself loads of hassle as the machine will run what you have currently with mimimal problems. However, if your plan is to drift/migrate toward Itanium then the other posts are more significant.

Mike.

[Steve Lewis]

Kieran,
First of all, I recommend that you take your hardware upgrade budget and use it upgrade Oracle sooner rather than later, in place, on the old platform to a version that is also supported on the latest hardware.
Then your options will be far more open and the next generation of hardware you get will be *much* faster.
The PA-RISC servers are not going to be sold much longer and the cost per PA cpu is way higher than for an IPF cpu. Get a comparative quotation for an rp3440 (PA) and compare it with an rx3600(IPF).
HP want you to upgrade and they really want to keep you as a customer, so the pricing for a newer model of server will be advantageous.
You won't be able to restore a recovery tape to an IPF server, but you will be able to use the same in-place oracle data storage.
As a pure back-end DB server running Oracle, your migration to IPF will be much easier than having to port an in-house application.

[Kieran Wareing]

Thanks Mike.

Second user would be okay. My company just wants to know all the options because I am having to write the DR plan from scratch. (no I can't believe we never had one before either!).

Knowing that the company wallet rarely sees the light of day, it could be years before they splash out on modern hardware to replace our old servers. But that's not to say we shouldn't use something more modern for DR if the solution is workable.

We have the option of hiring/borrowing something to test with so we can find out if the software works. Which brings me back to the 'what model' question.

Kieran

[Steve Lewis]

The other thing is, if you are running 2 instances of Oracle on the same server (why not split them?) and have just one DR server, then you will have to double the memory.
For the later versions, people are now talking about a minimum of 8Gb RAM for a single instance (ie 16Gb for 2 instances) - but memory in these boxes is expensive. So that raises a question about the possibility of combining your instances into one large one - i.e. the same Oracle version and not spending so much capital on memory. You will have to weigh the costs. Most organisations would rather spend the money on memory and keep two separate installations because its the least risk option.

[Kieran Wareing]

Thanks Steve
I agree with everything you say but there is no money. No-one in the company even gets a pay review this year!

DR is being forced on them in order to comply with a BSI standard (this is England by the way). So any money allocated will be the least they can get away with. It is very frustrating.

We are currently migrating the old Oracle 7 DB to Oracle 9 so that money at least is guaranteed. But it's not likely to be ready before the end of the year.

Sorry everyone I shouldn't be using this forum to complain about my employer!

Anyway, as long as I can give them some alternative server specifications then the ball is in their court.

Thanks
Kieran

[Steve Lewis]

Well in that case there are other options:
The first one is a specialist DR company like Sunguard availabililty services, who will quote you for a PA server that they already have in their data centre near Heathrow and some SAN storage. You just turn up with your tapes and do the recovery. You will also have to sort out the networking from them to your users.

The second (cheaper) alternative is to find a local company and do a back-scratching DR agreement with them. Your spare server for their use and their spare server for yours. Sounds easier than it probably is and a bit glib I know.

The third idea is to go commodity (ie PC) hardware and Linux for your DR database hosting. You would effectively be doing a migration for each DR but you may make it work if you can get away with setting set the service standards low enough - e.g. a regular export/unload of data instead of a RMAN backup or fbackup.

[Mike Shilladay]

Hi Kieran,

Do you have the specs (memory, disc) for your existing server? I am convinced you would be surprised by the costs. If the requirement was to run in the eventuality of a disaster albeit at a slower speed you may find that something like rp5470/4 way might be enough. (Max 16GB/4 CPU's at 550MHz)

Mike.

[Steven E. Protter]

Shalom Kieran,

Server model recommendations require some idea as to the size and activity of the database.

For Oracle without any application servers a 2 GB RAM is considered the minimum.

To run Oracle 8.1.7.4 which is out of support you can't go with the more powerful Itanium servers because they don't run HP-UX 11.11

An rp3440 with 2 GB of RAM is more than powerful enough to run up to intermediate activity Oracle databases in the 7/8 range.

SEP

[Kieran Wareing]

Hi
The current servers each have 3 CPUs and 4 Gb memory. The Oracle 7 and 8 DBs are each split across 4 x 18Gb FC disks. Although the 7 DB would fit on 1 x 18Gb and the 8 DB is just a little bigger.

Max users per day is < 200 but we could manage with less than half that number in a DR situation.

Kieran

[Mike Shilladay]

Hi Kieran,

If you are prepared to take a modest performance cut, an rp2450/2 way or A500/2 would probably do it, (rp2470 will also work but is more expensive as that is still sought after) with 4GB RAM (8GB MAX capacity), 2 x 18Gb internal drives for OS and some external drives for the application, SCSI drives will be cheaper.

I am thinking that a price of circa 4-5k pounds sterling. This solution will be 3U, but no tape is present.

If you want the same or better perfomance then the rp5470 will be best, but will cost more.

Mike

[Kieran Wareing]

Thanks Mike

The company expects a cut in performance so that isn't a problem.

At least now I have a starting point to work from.

Thanks again to everyone who contributed.

Regards
Kieran