Request for comments / design advice 5

[AnotherAlan]

Hi all,

I would like to seek some expert opinions on how best to arrange / re-arrange my exisiting architecture.

Currently:
Two seperate NBU environments, both on Solaris NBU v6.0 MP6.
1) One master with L25 tape library and one media also with L25 tape library.
Catalog size close to 50GB

2) One master with attached C4 tape library plus 5 media servers all with C4 tape libraries.
In addition to this, two of the media servers have 6TB SAN storage presented to each of them and this is configured using ZFS Pools.
Catalog size close to 75GB

Ideally I would like to provide a bit more resilience to the entire infrastructure by combining the hardware and creating an opportunity to failover to an alternative master server if required.
I'm really not sure how to do this though, especially what to do with the catalogs, or even if they can be combined.

All help appreciated.
Thanks
Al

 

[OzzMosiz]

This will require Symantec consultancy as merging Masters is not an easy thing to do!

 

[AnotherAlan]

Thanks OzzMosiz,

I was hoping to avoid that...

 

[OzzMosiz]

AnotherAlan, sorry dude, it is pretty complex stuff and you will want to ensure you have as little downtime/issues as possible.

 

[AnotherAlan]

No need to apologize. It's good to get some straightforward honest advice.
I suspected that it would not have been an easy task...best let Veritas take the heat if it goes wrong. ;-)

Thanks

 

[comtec17]

Symantec consulting has a catalog merge utility that they use internally to do this work and currently do not allow other consultants/partners do have access to it.

 

[RustyMajor]

For a failover Master, the best option is a stand-by server, i.e. a 'cold' server. You could do clustering, but it's a pain in the buttocks and you have to consider things like redundant pipes, etc. if you are going to do a geo-cluster.

For the environments, what is your retention and how many images? If you aren't keeping that much data for very long, you could just let it expire. You could also import it, which is the poor man's way of merging catalogs .

 

[AnotherAlan]

Thanks RustyMajor and comtec17,

I have monthly backups with infinite retention periods so will need to keep the catalogs active for some time yet.

I've read a bit about the catalog merge utility. Doesn;t sound that tough...;-)

Al

 

[Skywalker1957]

I'd drink some coffee before putting all my faith in Symantec. They aren't what they used to be when it was Veritas. Just warning you. See my post on

http://www.tek-tips.com/viewthread.cfm?qid=1517848&page=1

 

[AnotherAlan]

Thanks Luke, there is some good advice in that link.

 

[SANEngInCO]

Alan,
Like everyone is saying, merging catalogs is pretty tricky. I know people who have done it, but those cases are very rare. Consultancy is a good idea at least for that portion of it.

As for some ideas around your architecture, here are some thoughts. To gain resiliency, I would recommend using Veritas VCS and combine your two masters (as long as they are the same type of servers). In my old environment, I combined a Windows Master and a Solaris Master into a single Solaris environment. Then I pushed that new environment into a VCS active/passive cluster. They were both SUN V440's, but I had twice the CPUs and memory in the active node since the passive was just for failures and updates (basically short-term usage). It ran brilliantly for years (I think it still runs today 2 years after I left).

I would also suggest consolidating your libraries into a larger more redundant library (STK L700 for example) that runs via Fibre Channel. This will give you lots of growth potential and faster throughput verses things like SCSI attachment. This would also allow you to do sharing to multiple media servers via SSO. I had consolidated a SCSI STK L40 and ADIC Scalar 1k into a fibre channel STK L700e with my previous environment and it worked great. Speeds increased, windows shrank and backup failures almost completely went away.

As for your storage (2 x 6TB), are you using them as DSUs or DSSUs? If not, that is what I would highly recommend. I had 8 DSU mounts in my old clustered environment that were used for the most critical backups (Oracle, SQL, Exchange, etc) that needed to be done quickly and with the most resiliency. We were then able to duplicate the backups via Vault to the library and then send those tapes off-site. It also allowed us to keep many days of those backups on hand and on faster media for restores.

All said and done, what you are dealing with is very similar to what I did 4+ years ago. With the changes I mentioned, the environment became super stable and the reliability of the backups increased exponentially.

Good luck with your environment. It's an exciting (yet tiring) thing to do!

------------------------------------------------------------------------------------------------------------------
"Only two things are infinite, the universe and human stupidity, and I'm not sure about the former."
Albert Einstein

 

[AnotherAlan]

Thanks SANEnginCO,

I am going to look at a possible active/passive cluster next year, budgets being approved of course!
With regards the DSU and DSSU, I have a mixture, 4TB from each has been assigned as straight NFS mounts to Oracle servers for RMAN backups and the remainder have been configured as DSSU's.
I agree with the consolidation to a L700. I've been after one of those for a couple of years, but management have only just started to see the light. Again, maybe next year.

I appreciate all the advice that has been given on this post so far, thanks all and enjoy the holidays.

Alan