data block corruption + replication

ORACLE 8.0.5
+Advanced replication
NT 4.0 SP 5

Monday after a reboot of the NT machine we experienced a data block corrupted error. The block in question was block#2 in file#1. This is where the SYSTEM rollback segment lives. I was still able to pull dumps from most users. Then I tried to shutdown + restart but the database would not go beyond mount.

(1) AVOID
    There are no indications about an OS error. Both the NT logs and     the Compaq logs show nothing exceptional. So how could such a     thing happen. The disk is BTW a RAID5 array. Seems like an Oracle     provblem to me. What do you think.

(2) EXPORTING MASTER OR SNAPSHOT
    The machine in question is the master of 22 updatable snapshots. I     was able to recover the master fairly ealisy from my dumps. But it     did not know much about replication afterwards. No master groups     no nothing.

    I believe that the knowledge about replication is located in the     SYS/SYSTEM schema. This knowledge was created by calls to the     replication api, when the master was initially set up. While     oracle faithfully puts CREATE TABLE statements into a dump in     order to reconstruct the knowledge about tables (which also lives     in SYS/SYSTEM) it does not do so for replication objects     (repgroups ...). Is this true ? If this is true, even a full     database export would not help.

    I noticed thoufh, that for a snapshot there is a CREATE SNAPSHOT     statement in the dump.

(3) SINGLE POINT OF FAILURE
    With this setup (one master, 22 snapshots) I have an ugly single     point of failure. My snapshots won't do much intelligent work     anymore, when the master is down, even though they are physically     alive. Is there a way around this ?

(4) SYNCING
    When I recover the master to a point in time in the past, it will     be out of sync with the snapshots and the system as a whole will     be inconsistent. I am currently rebuilding all my snapshots     in order to get them in sync with the master. This takes apx 4     hours per snapshot and does not run fully automated. Is there a     smarter way of doing this ?      Received on Thu Apr 05 2001 - 02:23:27 CDT