Environment Info:
Oracle Version: 7.3.4
Platform: Solaris 5.6
HDS Platform: HDS 9900
HDS MicroCode Versions: HDS#1=01-18-09-00/00
HDS MicroCode Versions: HDS#3=01-16-40-00/00
HDS CCI version: 01-09

Hello Everyone Again,

I am having some challenges at the moment getting Oracle to recover to a specific log file after a HDS Shadow Image split. I am taking a hot backup of the source server, doing 3 log switches, splitting the mirror and using the oracle recover until cancel command on the target server.

I would like to be able to recover to a specific log file on the target rather than recover automatic. I do notice that the last log being split is actually a partial log. That is the archive log on the source is larger than the same archive log on the target. However, since I don't mind to recover to an earlier point I don't necessarily care about the last log.

My problem is this: Whenever I issue "CANCEL", Oracle will continue to prompt for the next log. It refuses to bring up the database consistently. I will have to apply each log, up until the last archive log. When it reaches the last archive log, it will complain that the file is corrupted. I am able to recover auto fine, but I actually need to recover to a specific earlier log file, not the last one.

I always thought when I issue CANCEL, Oracle should throw away the rest of the transactions after the log that I cancelled? For some reason it is not able to recover consistently.

We also split the datafiles earlier than the log files. Before each split we will issue a few log switches. Another thing that I noticed was that if "current" is seq#100, last redolog file seen in source server must be 99. Yet we keep having only #96, #97 available in the target.

So basically a few questions from this scenario:

(1) How does recover until cancel really work? Why doesn't oracle let me cancel?
(2) Is there something wrong with the way we split the HDS? Could it be the split occurs before the disk is synced properly or something?
(3) Does anyone else have their own HDS split scripts / sequence of events / steps which they use for a consistent Oracle Hot Backup that I can compare with our own?

I think I am too close to the problem to think clearly any more, and I am looking forward to hearing some fresh ideas for all of you!

Thanks & Regards