Implementing continental cluster with RAC 9i with EMC SRDF synchronization technology

[Oleg_30]

We have a two sites environment, each holding three servres equiped with MC/Service Guard OPS edition with RAC 9i (running on two of the three servers). The storage they use is EMC symetrix. The two sides synchronize data using EMC SRDF mechanism. according to continental cluster release notes, RAC with continental cluster is supported only on HP continues access XP. Do you know a way we can build the above mentioned cluster without using the XP synchronization technology.

[melvyn burnard]

As you have noted, Oracle RAC is only supported in a CC when using XP arrays.
I would guess you might be able to buy the CC product and then hack it, but is that really what you want to do if this database is so important to you?

[Oleg_30]

Of course my main concern is the consistency and availability of th DBs, but I haven't chose the storage system that affects my whole organization which most porion of it is non HP-UX by the compatibilty with HP continental cluster. And I won't throw away the machines which I already have, I just wanted to know if there is some one who implemented, or know of way implementing this cluster with EMC storage. My other option is not using automated cluster solution but writing scripts my self which will implement 99% of it's functionality and I wanted know if there is other way that's all.

[Michael Steele_2]

There are other HP users using EMC SRDF in MC/SG over large geographical distances between data centers and I've heard them asking about RAC as well. But they've dismissed it because they???ve got a 4 hour latency between BCV's and a database syncing issue.

RAC is a utility designed to load balance network transactions in a real time environment?

Do you want to get both data centers (* which are spread across how many miles? Hundreds? *) up simultaneously and each handling 'real time' Oracle transactions with RAC and expect a performance improvement and still keep both EMC disk arrays sync???d up?

How long does it take for the alternate BCV to sync up with the primary? How many minutes? Hours?

[Chris Vail]

We'll be building the same kind of thing here. But as of the moment, we're not using Oracle RAC. We have two EMC 8830's (with 69.8 TB each) about 45 miles apart. Right now they're running in adaptive copy mode, rather than synchronously as planned. We have serious concerns about bandwidth between the two sites as well as the 8830's internal bandwidth. Right now, the second, failover site really isn't keeping up with the primary.
We have a project in place to move the HP systems to Metro Cluster, rather than Continental Cluster. But this is just now kicking off.
I think the secret to making both of our projects work is to get EMC to sign a Service Level Agreement for you. That makes THEM responsible for making it all work. That'll jack up the prices, of course, but they'll guarantee the performance.


Chris

[Oleg_30]

Michael hi,
I have RAC in both sites, of course the RAC is insite, and the replication of the data will be probably synchronic. Assuming i have enough bandwidth.

[Rita C Workman]

Well I'm no Oracle guru...so I'm really not familiar with RAC9i.

That said...we just implemented a CC with EMC storage using SRDF. We're an Oracle shop with databases of varying sizes. Our DR site, for now, is only about 15-20 miles down the road. To make the DR/CC project more interesting, they wanted it bi-directional. So we have production db's on both sides of our CC, each failing to the other site. Site 1 are older K boxes, the DR site is new RP8400's-all are 11i/64 bit running MC/SG 11.14. Each site has an EMC Symm8530. One site a SAN...the other host-based. We even threw PowerPath on everything.

For our largest database, we found after the initial db sync that we could run an adaptive copy to split our r1/bcv's from the primary disks...sync them over to our DR site's r2 disks about every 2 hrs (only takes about 20-30 minutes to refresh).
Now we do adaptive copy for everything 'data'...but our ORACLE LOGS we KEEP IN SYNC. So we set up a diskgroup in sync mode and made sure a copy of necessary ora_logs are there...control,archive,redo.
We've tested and so far the DBA's have been able to get everything up on the other site....no kidding - in minutes (they checked db and declared it sound everytime so far) !
Just to get a little crazy we added a Gold BCV behind the r2, scheduled nightly to sync it back to the r2 and then split it off after and bring up this GoldCopy of the db on a box at the DR site (in read only & outside of the MC/SG cluster). We use it for backups and the folks here run their really big report demands against it...thus freeing up the old K box production server. A true win/win for everyone.

We have smaller db's that we keep in sync mode, since the traffic and changes are so small...no real impact.
When these fail we simply do a split and then have them run off the r2 at the other site.

Now what your refresh time takes will depend on how much data there is. And as you can see...we use more than a few disks. But no OracleRAC.

Just how we did it....
Rgrds,
Rita

[Oleg_30]

Thank you very much Rita, thats pretty much they way we plan to do that, even bi-directional and BCV.
We haven't decide the way we'll sync the data, the logs will be synced of course, but the most important is RAC, because we need the DB's highly available and no one can answer how can I synchronize it over SRDF. Its not oracle issue but OS/SG/SRDF issue, and I desperatly need someone who've done it or heard about something like that