RDF System Management Manual

Triple Contingency
HP NonStop RDF System Management Manual—524388-003
10-5
The COPYAUDIT Command
RETAINCOUNT is set to three, the purger process can only purge AA000010 (it must
keep AA000011 and AA000012 on disk). Thus, as long as the RTD times of the
extractors on the two backup systems are less than 36 hours apart, the triple
contingency protocol will work successfully.
Similarly, if you believe the two receiver processes will never be more than 48 hours
apart in their RDF processing and your image trail file sizes are such that
approximately 20 rollovers occur every 24 hours, then you should set the
RETAINCOUNT to 40 on both backup systems.
You set the RETAINCOUNT parameter by issuing the following command:
SET PURGER RETAINCOUNT num
where num is a number within the range 2 through 5000. The default is 2.
You can alter the RETAINCOUNT only when RDF is stopped.
You alter the RETAINCOUNT parameter by issuing the following command:
ALTER PURGER RETAINCOUNT num
where num is once again within the range 2 through 5000. (Before entering this
command, however, you must first stop RDF.)
The COPYAUDIT Command
If the primary system fails, you must execute two takeovers: one on each backup
system. Upon successful completion of both takeovers (signalled by a 724 message in
the EMS event log of each backup system), the databases on the two backup systems
will almost assuredly be different: one of the systems will have been further ahead of
the other in its RDF processing when the failure occurred.
The COPYAUDIT command copies missing audit information from the backup system
that was further ahead in its RDF processing (had the most amount of audit data) to
the system that was further behind (had the least amount of audit data).
Upon successful completion of the two takeovers, examine the EMS event log on both
backup systems for a 735 message. That message, which follows the 724 message in
the log, specifies the last position in the MAT that was seen by the receiver process.
Compare the MAT positions in the two 735 messages and determine which of the two
systems was further behind in its RDF processing when the failure occurred (that is,
which system had received the least amount of audit data from the extractor by the
time the primary system was lost). Then issue a COPYAUDIT command on that
system, specifying the name of the other backup system and its RDF control
subvolume.
The syntax of the COPYAUDIT command is as follows:
COPYAUDIT, REMOTESYS sys, REMOTECONTROLSUBVOL subvol