Zlatko, ZK>The only way to "break the mirror" and access both copies in AIX known to ZK>me is to pull the disk and put it into another machine importing the VG. ZK>Otherwise the mirror is still transparent to you through AIX APIs and ZK>tools.
Not so... The sbom_backup script does some pretty hairy things, but the end result is a broken mirror. One part of the mirror can be used as normal but the broken portion (which are mounted as phantom filesystems under a different mountpoint) cannot be written to, but does offer a point-in-time view of the mirror at the time it was broken. What sbom_backup does is (this is the description from the script itself)... # This program pulls the wool over LVM's eyes by taking a disk already # defined to a volume group and making LVM think the original disk is # temporarily missing and a new disk has been added to the system. # This allows us to create a new volume group on the disk, build the # logical volumes using the PP maps of the originals, then mounting and # backing up the data in the filesystems. The copies cannot be written to at any point. Filesystems are mounted read-only. It is an impressive, and scary piece of code. The bit where it writes a new PVID to the disks worries me the most... ZK>In Windows NT you can break the mirror of a partition and late the copy can ZK>be assigned another letter. At least I am not familiar does such method ZK>exist in AIX. I can tell you more details of how this is accomplished in sbom_backup if you want. Also, I suggest you have a look at the redbook mentioned in the post by Steve Harris. chlvcopy in 4.3 appears to be able to do this kind of thing (though the "Use at your own risk" warnings do not inspire confidence!). ZK>Are you files so intensively accessed (if yes by what application) and ZK>aren't you satisfied by "shared static" and "shared dynamic" serialization ZK>options? Hot backups are not an option. Please see my reply to Miles for more detail on our situation. Regards, Steve Greatbanks --
