Disaster Recovery in Windows 2003/Cluster, SQL 2000 and R3

[Former Member]

Hi,

Can someone share experience/knowledge of disaster recovery scenarios in MSCS/SQL Server/SAP. One of our customer has R3/SQL Server2000/Win 2003 (Cluster).

We would like to evaluate best possible options for the Disaster Recovery which are supported by SAP.

We have thought about

1. Log shipping

2. Standby Database

3. Restore backup on new cluster

4. Homogeneous System copy.

We do not want to go for first two and would like to explore on 3rd and 4th option.

Any links to documents/blogs will be helpful.

Thanks,

Manoj

[markus_doehr2 · Accepted solution]

> We would like to evaluate best possible options for the Disaster Recovery which are supported by SAP.

> 1. Log shipping

> 2. Standby Database

> 3. Restore backup on new cluster

> 4. Homogeneous System copy.

>

> We do not want to go for first two and would like to explore on 3rd and 4th option.

3 and 4 are technically the same, you install a new system and restore the database.

The question is: what is the requirement from you client (how much downtime etc.)?

Markus

[Former Member · Accepted solution]

Markus,

Thanks for your reply.

3 and 4 are technically the same, you install a new system and restore the database.

I am confused. Option 3 will be restoring backup and 4 will be sapinst. Isn't it? Are both options supported by SAP?

Is there a SAP standard documentation for building cluster from scratch and build SAP system from backup or sapinst for DR?

I am sure there will be installation document if it is a fresh installation. But not sure if there is one for DR.

Manoj

[markus_doehr2 · Accepted solution]

> I am confused. Option 3 will be restoring backup

Yes - but what will you restore? Everything? If you're running on a cluster it's unlikely that both nodes will fail at the same time so there is still one node that can and will run the software, no?

> and 4 will be sapinst. Isn't it? Are both options supported by SAP?

Yes.

> Is there a SAP standard documentation for building cluster from scratch and build SAP system from backup or sapinst for DR?

The standard installation documentation cover a cluster installation.

> I am sure there will be installation document if it is a fresh installation. But not sure if there is one for DR.

If you have a cluster then you have a high availability already. If a node fails, you will "just" reinstall that node and put it back into the cluster.

What kind of DR scenario are you thinking about?

Markus

[Former Member · Accepted solution]

I will check the standard documentation. but will standard documentation contain restore backup for DR? We are planning to simulate the situation like complete production servers/SAN are flooded and only tapes are available offsite. We may not get last minute data but we want confidence we can build the servers/cluster from tapes. So plenty of time is available from management to simulate this situation.

[markus_doehr2 · Accepted solution]

> I will check the standard documentation. but will standard documentation contain restore backup for DR?

There is no special documentation to set up a system from scratch or to restore the system from backup. Why should it?

> We are planning to simulate the situation like complete production servers/SAN are flooded and only tapes are available offsite.

Ok. Then reinstall the OS, backup software and restore the cluster.

I'm still not sure for what kind of documentation you're looking.

Markus

[Former Member · Accepted solution]

Then reinstall the OS, backup software and restore the cluster.

Do you mean install OS on both nodes, SQL Server, setup clustering , backup software and then restore the backup? This is what I am thinking but I am not clear about this in cluster scenario. That is why I was looking for any standard documentation for this situation. In the meantime, I will go through the standard SAP installation documentation in cluster environment.

[markus_doehr2 · Accepted solution]

> Do you mean install OS on both nodes, SQL Server, setup clustering , backup software and then restore the backup? This is what I am thinking but I am not clear about this in cluster scenario. That is why I was looking for any standard documentation for this situation.

If you assume for a DR scenario that everything is destroyed then the re-setup is basically a new installation (as if there was nothing) with the slight difference to use the database backup from the tape as opposed to loading the database from the DVDs.

> In the meantime, I will go through the standard SAP installation documentation in cluster environment.

Yes, that's what I mean.

Markus

[Former Member · Accepted solution]

DR generally would be at an alternate location

It will be a totally new installation but what you can make sure is that you get from both the nodes

1. if you can install and make standard OS image which is backed up and then just use that to restore

2. Snapshots of your registry

3. Docs for installation and post processing

4. etc...

You need to get all of the above from an Offsite location so you can simutlate the process.

Also gives you an option to test the offsite procedure

Pravin

[Former Member · Accepted solution]

Is it supported by SAP to take golden image/point in time snapshot of MS cluster and then restore on another set of nodes? Any documentation/link will be helpful.

I am looking for something similar like HP_UX ignite. I used this for various applications and worked fine with 2 steps. 1. Restore Ignite on new system 2. Restore backup at the same location on new system.

I am still confused about SAP DR and trying to find out what is the best possible solution which takes less time and works in all the situations.

[Former Member · Accepted solution]

Markus,

If you assume for a DR scenario that everything is destroyed then the re-setup is basically a new installation (as if there was nothing) with the slight difference to use the database backup from the tape as opposed to loading the database from the DVDs.

This is where I have gray areas. See if I understood this.

1. Install SAP in MSCS on new hardware , configure and create datafiles/log files as per the previous setup

2. Finally replace these files from the database backup.

Will this work?

Manoj

[Former Member · Accepted solution]

Is it supported by SAP ??

Hmm, Never even given that a thought as this was a restore from OS and we are nto changing any SID.

Have participated in a few DR tests and never thought fo SAP support to be an issue.

Fastest way and easiest way to get it all set up would be

Restore of an Image of OS with Patches and then the restore of the registry sanpshot and then restore of the database.

Pravin

[Former Member · Accepted solution]

<Hmm, Never even given that a thought as this was a restore from OS and we are nto changing any SID.

Have participated in a few DR tests and never thought fo SAP support to be an issue.>

I do not want to change SID in this case because simulation will happen on standalone setup and in case of disaster we need to same SID.

<Restore of an Image of OS with Patches and then the restore of the registry sanpshot and then restore of the database.>

Not sure how this will work in cluster set-up. But I think we are thinking in the same direction.

[Former Member · Accepted solution]

Agree that no SID change will be need as this will be a DR

Would do the restore as if they are distributed i.e

CI on one and DB on the other

And then would go through the cluster setup.

Have not done MS Cluster but would think that would not take too long as long it is documented.

Pravin

[Former Member · Accepted solution]

Hello, I'm doing a "Plant ECC6 to do a test disatre / Recovery.

During the installation, I would use the same <sidadm> existing as it is in the PRD.

If I tell you to install it during the same <sidadm> it will overwrite what already exists, or do I just enter the password of existing <sidadm>?

Regards

Guilhermino

[Former Member]

Hi Manoj,

For our clustered system we use two logical steps for the recovery (in a DR 'data center blown away' scenario). First we restore the two cluster nodes using disk images of those nodes (using a software package like Acronis). Then we create new storage from the SAN side, assign that storage to two nodes, restore that SAN storage, and/or restore from a DB backup.

I don't know if that is helpful but that is the scenario we use for a restore.

J. Haynes

[Former Member]

We use our QA system as our DR point... as our QA environment is in another location and it is an exact replica of what we would need. Both QA and PRD are a Win 2003 EE R2 x64 2 node cluster. We then could just restore the database from tape and then do a SID change. The do some fun DNS stuff and we're done. Basically it just allows you to get around having to reinstall everything.

[Former Member]

I am still not clear on how this will work in MSCS environment. Do I need to take backup from both nodes and restore on both nodes and then DB? At what stage I should include nodes in cluster?

[markus_doehr2]

The question is: what is your goal and how much time do you have?

Are you intending to build a DR cluster or do you just want to build up a (single) system so people can work?

Markus

[Former Member]

The question is: what is your goal and how much time do you have?

Prepare exact replica image of production server from backup. We are in simulation stage so we have enough time to set-up the process. But in case of disaster we want to spend as much as less possible time on the recovery.

Are you intending to build a DR cluster or do you just want to build up a (single) system so people can work?

We are planning to build as is scenario from backups/tapes/ASR disk (Whatever required) but if this does not work then we may need to opt for single system in DR situation. Any suggestions are most welcome.