Oracle Active/Passive move to Oracle Parallel Server OPS!

[Ghulam Shabir Mallah_1]

Hi Friends,

I need yours adivce.

I have Oracle 8.1.7 Enterprise Edition (64-bit)
on HP-UX 11.00 (64-bit). System is currently live and I don't have any Test or Staging server where I can do this stuff. I am not in position to take any risk with live system.

Customer are asking me to move current system for OPS.

Can I move easily current system for OPS.

One is my friend adviced me for RAC but in my understading that RAC is difficuilt on this stage. Becuase for RAC I need to format all storage harddisks as RAW and that thing I can't affoard. Again this is production live system and don't have TEST or Staging system.

Thanks

Ghulam Shabir Mallah
GSM

[Brian Crabtree · Accepted solution]

Ghulam,

You are going to have to do the same thing (Raw disks) for OPS as well as RAC. This is required so that the disks can be shared between two systems at the same time.

I would highly recommend RAC over OPS. There are a number of configuration requirements for OPS, and it is generally better if you separate applications completely in an OPS enviroment. RAC has come a long way in block tranferring between instances.

Unless you have worked with Raw devices in the past, I wouldn't recommend moving to RAC or OPS though. There is a bit more administration overhead associated with it, and it is more complex than cooked filesystems when it comes to space planning and additions.

If you decide to move forward with it, the best option you have is to backup and export the current database, build a new instance on raw devices, and import. You really should use a test server before attempting this, so you know what problems will come up. RAC is relatively simple compared to OPS, but does take a bit to get running, especially the first time.

Thanks,

Brian

[Indira Aramandla]

Hi Ghulam,

Here are few suggestions.

You will have to have Raw devises for OPS as well as RAC. Raw devices are required with Oracle OPS mainly because of the disk sharing that has to happen at the OS level. OPS requires the shared file system for all the instances to share the files in the filesystem. This capability is provided by Raw Device

The following are necessary if you want to setup OPS.

1. RAW File System
2. MC/Lock Manager and MC/Service Guard
3. Planning for OPS configuration (High Availability), n-node clustering, basic OPS configuration (primary and secondary)???.etc.

If you do not have the required MC service guard installed, the installer will not give you the ops option. When you install the required OS dependent software, you need to go thru the installer and install the ops option and relink the oracle.

Mostly People use OPS for Load Balancing and Failover, in both cases you need to have the 2nd instance up and running.

You need to configure all raw devices for the OPS database creation, I mean all data files, redolog files, control files. Ensure that you create a atleast 50MB raw device for control file. People had problems with raw devices and control file sizing. And remember to define the raw device 1 MB > the actual size of the datafile as you cannot increase the raw device. You cannot extend the raw devices logically. Create additional raw devices and keep them as standby. If for some reason a TBS goes out of space then you can add the data file immediately.

As you said that you do not have a test environment, it would be prefered if an initial planning for OPS was done and have all the pre-requirements for OPS available before OPS setup.

IA

[Indira Aramandla]

Hi Ghulam,

Refer to the document for step by step instructions to setup RAC on HP_UX.




IA

[Yogeeraj_1]

hi,

to add to the above replies...

OPS is designed for highly available systems. You can protect your disks with raid. you can protect your network via redundancies. If the computer with your database fails however -- its all over. With OPS -- many computers in a cluster have the database open -- there is no single point of failure.

OPS is designed for highly scalable systems. Need more resources, drop in a new box.

RAC is a new more robust OPS.

OPS is not RAC, you have to design your applications for OPS!! (and we always said that - so what you had was a poorly designed OLTP app on OPS).

The thing with RAC is that a bad programming choice that affects you marginally on a single instance may will be magnified on RAC (or vice versa). So something that runs "ok" in a single instance may fall apart on RAC (and vice versa)

With RAC -- you do not have to physically design to scale up in a cluster -- very very different architecture.

RAC allows you to use more then one computer in a cluster to make sure the failure of a single computer (n-1 computers actually) doesn't bring down the database. By running more then one instance of oracle on different computers -- all accessing the SAME database -- we can increase (greatly) the availability of the system

If you had a 9i databases, the dataguard features would have been really great to implement....

regards
Yogeeraj

[Ghulam Shabir Mallah_1]

Dear Friends,

Thanks a lot for providing me lot of information on my subjected task. I am proud of you all.

Thanks

GSM