HP NonStop RDF, Руководство по управлению

Страница: 273 / 478

It is strongly recommended, however, that the various RDF process priorities be identical on
both backup systems so that the performance of the two systems is approximately the same.
WARNING!
If the two backup systems are configured differently from one another in any
important regard, the triple contingency feature will not work when you need it, and there will
be no advance warning to that effect. Prior to a primary system failure, the two backup systems
run independently of one another and no cross-checking whatsoever is performed to verify that
they are configured compatibly for triple contingency recovery. To guarantee that the two backup
systems are configured compatibly, it is strongly recommended that you FUP DUP the OBEY
command file that you use for initializing and configuring one RDF configuration, change the
suffix character and the backup system in the INITIALIZE RDF command (and perhaps some
of the process names), and then use that same OBEY command file to initialize and configure
the second RDF configuration.
The RETAINCOUNT Configuration Parameter
The purger RETAINCOUNT parameter specifies how many image trail files (including the one
currently in use) must be retained on disk for each image trail. The default value for this parameter
is two.
This parameter is important because if you lose the primary system, the triple contingency
protocol will work only if one of the backup systems has retained all of the audit records that
the other is missing.
For example, assume that you have lost the original primary system (\A), you have successfully
completed a takeover on both backup systems (\B and \C), and the MAT positions displayed
by the respective 735 messages are:
\B: 735 LAST MAT POSITION: Sno 10, RBA 100500000
\C: 735 LAST MAT POSITION: Sno 10, RBA 100000000
500 kilobytes of audit records is missing on \C.
Suppose that the image trail files are relatively small, such that the audit record at MAT 10, RBA
100000010 was placed at the start of image trail file AA000025 on \B. If the purger on \B is
allowed to purge AA000025 before the takeovers occur, the triple contingency protocol will fail
because \C is missing some of the purged audit records (Sno 10, RBA 100000010 through Sno
10, RBA 100500000).
The RETAINCOUNT parameter is designed to prevent such a situation, although it is up to you
to set this value correctly.
You must determine how much time disparity to allow for in the event that one receiver falls
behind the other. Such a disparity would occur, for example, if the communications lines between
the primary system and one of the backup systems were to go down for some period of time.
The RETAINCOUNT parameter must be such that no image trail files that might be needed for
triple contingency are ever purged.
The best way to do determine the appropriate RETAINCOUNT value is to pick an acceptable
time differential such as 24 hours, 36 hours, or 48 hours, determine how many image trail rollovers
typically occur within that amount of time, and then set the RETAINCOUNT parameter to that
number of files.
For example, if you believe the two receiver processes will never be more than 36 hours apart in
their RDF processing and your image trail file sizes are such that rollovers occur only once every
24 hours, then you would be safe specifying a RETAINCOUNT of three for both backup systems.
In that situation, the purger process on both backup systems will always keep at least three image
trail files on disk (the one the receiver is currently writing to and the previous two). Assume, on
the backup system which is further ahead in its RDF processing, that files AA000010, AA000011,
and AA000012 are on disk, the receiver rolls over to file AA000013, and all updaters have just
begun reading file AA000013. Files AA000010 through AA000012 might be considered expendable
The RETAINCOUNT Configuration Parameter 273