RDF System Management Manual for H-Series RVUs (RDF 1.8)

If you initialize RDF at the shutdown point at Step d, you should restore on the backup
system a copy of the primary system database taken at Step d. The databases would not be
synchronized if the database at Step b was restored to the backup system.
If you initialize RDF to the timestamp corresponding to Step b, you should restore on the
backup system a copy of the primary system database taken at Step b.
• Initialize RDF at the most recent TMF shutdown point. If you initialize RDF at an earlier
shutdown point, RDF operations will start at that point but will shut down when the next
TMF shutdown point is reached. In this case, you must restart RDF quickly so that operations
on the backup system do not fall behind those on the primary system. If you choose to
initialize RDF at a TMF shutdown point that is not the most recent, watch the RDF event
messages for the RDF shutdown message and then restart RDF immediately.
• If you include the TIMESTAMP option in the INITIALIZE RDF command, use the following
guidelines to determine when you must restore the backup database:
— If you are going to start RDF with UPDATE ON, restore the database to the backup
system before you start RDF.
— If you are going to start RDF with UPDATE OFF, you do not have to restore the database.
However, if the need for an RDF takeover arises, you must then restore the database
on the backup system before you issue the TAKEOVER command.
• If you include the TIMESTAMP, INITTIME, or SYNCHDBTIME options in the INITIALIZE
RDF command, it is suggested that all MAT files are present from the current file to the file
containing the TMF shutdown record or the commit/abort record with a timestamp less than
the specified timestamp. If RDFCOM tries to open a nonexistent file, RDFCOM can trigger
restoration of the missing audit-trail files from tape or disk based on your response to a
prompt message. Therefore, not every file in the MAT must be present, from the initial file
to the current file. For example, if the current file is AA000010 and the shutdown record is
in AA000009, only files AA000009 and AA000010 need be present.
• In any event, if you plan to enable updating on the backup system as part of the new
configuration, ensure that the primary and backup databases are logically identical before
entering the INITIALIZE RDF command. For more information about database
synchronization, see “Understanding Database States” (page 147).
Examples
The following INITIALIZE RDF command, issued on the primary system, \LON, initializes the
subsystem to 2:30 pm, January 12, 2004:
]INITIALIZE RDF, BACKUPSYSTEM \CHI, TIMESTAMP 12JAN2004 14:30
Do you wish to proceed? [Y/N] Y
Please wait while RDF searches for the specified timestamp.
TMF shutdown at 12JAN2004 14:30 has been found. RDF will
start at RBA: 376275 MAT file: $AUDIT.ZTMFAT.AA000414
Do you still wish to start at this point? [Y/N] Y
The following INITIALIZE RDF command, issued on the primary system \LON after TMF was
stopped, deleted, and reconfigured, initializes RDF at once, without prompting you to confirm
your request:
]INITIALIZE RDF, BACKUPSYSTEM \CHI, SUFFIX 2 !
In the first example, the RDF control subvolume is implicitly named LON while in the second
example it is explicitly named LON2.
LOCKSTEP
You issue the LOCKSTEP command on the backup system when you lose your primary system
and decide to take over operations on the backup system. The LOCKSTEP command starts an
extractor on the specified standby system, and that extractor sends audit data from the audit
202 Entering RDFCOM Commands