Best Practices for CMOD Multiplatform Disaster Recovery

[JJeffrey]

It has finally been decided that CMOD in our shop is a critical Tier 1 application and, in the event of the loss of our data center, it must be recovered within 24 hours, including documents. The IBM documentation has the standard "keep a current copy of the software to install....."

Does anyone use High Availability or mirroring so that a switch can be flipped and CMOD is up and running at a hot site?  What are true best practices for recovering CMOD in the event of a disaster?  We are running CMOD 8.5.0.6 on RHEL, DB2 9.7 with TSM 5.5 on our old AIX servers.

[Alessandro Perucchi]

Hello JJeffrey,

might I suggest these rebooks/redpaper:

http://www.redbooks.ibm.com/abstracts/sg246444.html?Open
http://www.redbooks.ibm.com/abstracts/tips0517.html?Open
http://www.redbooks.ibm.com/abstracts/tips0518.html?Open

this will give you an overview of some practise on the field.
It depends also on you infrastructure, etc...

There are 2 parts to be aware:

- Library Server
- Object Server


For the library server, you have your database, and this is really to see what strategy you are using for backup/restore, high availability.

For the Object Server, you have CMOD Cache, and TSM.
I must say that everything that is in the cache, must be in TSM, or quite quickly "migrated" from cache to TSM.

Then the TSM part, you will need to use the built-in functionality of TSM, in order to create copy pools to have several copies of your documents in several locations.
There you can "play" with that concept a lot, it will just stop with how much money and how far you want to go.
We have some customer that have 2 pools, one in each remote location, and one location has in addition a 3rd pool with tapes. So in case that you cannot retrieve the disks, and both sites are down, then you could theoretically restore the tapes down to disk.
It might for that scenario take more than 24 hours, depending on the tapes, etc...

Well the possibility are endless, it just depends on what you are willing to create, what the different team (network/storage/unix/dba/...) can do, what is your budget, etc...

I hope to have given you a little overview on the topic. Which is a very broad topic in itself.

Sincerely yours,
Alessandro
Since you have TSM,

[Justin Derrick]

Unfortunately, the Redbooks that Alessandro linked to are 8 years old, and don't really represent the state of the art in DR for CMOD anymore.

There *is* an updated CMOD Redbook coming out, but I haven't seen it yet -- so I'm not sure what exactly is in it.

For critical CMOD installations, I recommend an active-active scenario where both servers are independent copies of each other.  This allows for load-balancing, and in the case of a failure, 50% of your users never know there was an outage, and the other 50% simply reconnect to the server that is still alive moments later.

For systems that can be down for 24h, I recommend you use your regular enterprise backup solution, restore it into a new server, and then have some DR planning for TSM separately.  (This usually involves an offsite copy of all TSM data -- usually provided by server-to-server backups across the WAN.)