RDF/IMP, IMPX, and ZLT System Management Manual

Managing RDF
HP NonStop RDF/IMP, IMPX, and ZLT System Management Manual—524388-002
5-8
Processor Failures
Updater Failure
If an updater process fails, RDF aborts. A subsequent START RDF command restarts
this process without requiring database resynchronization. To support restartability,
however, the updaters use a different mechanism than the extractor or receiver: the
updaters rely entirely on context saving rather than checkpointing. For this reason, if
the backup member of an updater process pair takes over because the CPU of the
primary member failed, the backup updater might have to start at an earlier point in the
image trail and require several minutes to reach the point where the primary process
was positioned when the CPU failed.
When a primary CPU on which an updater process is running goes down and then
comes back up, RDF does not switch the updater to run on the reactivated primary
CPU. Instead, once the backup updater takes over, it becomes (and remains) the new
primary process. If you stop and then restart updating, however, the original CPU
configuration for this updater process is restored.
Purger Failure
If a purger process fails, RDF aborts. If the primary process fails and the backup
process takes over, the current PURGETIME interval is aborted and a new one is
started.
RDFNET Failure
If the RDFNET process fails, RDF aborts. If the primary process fails, the backup
process takes over.
For further important information, see Section 13, Network Transactions.
RDF State Transition Failure
Periods during which the RDF updaters, or RDF itself, is either starting or stopping are
known as RDF state transitions. In rare instances, when a primary CPU fails for an
RDF process during execution of a STOP RDF or STOP UPDATE command, not all
RDF processes complete the state transition properly.
To minimize the chance of encountering this kind of failure, avoid CPU reloads during
RDF state transitions. Furthermore, if a CPU failure does occur during a state
transition, carefully review the EMS event log for signs of incorrect behavior. If the
failure occurred while RDF or the updating facility was stopping, check the Process
Pair Directory (PPD) to ensure that the appropriate RDF processes all have stopped; if
they have not, you must stop them manually.
If a state transition failure occurs during execution of a STOP RDF command and the
operation appears to be stalled, manually stop all of the RDF processes by issuing the
following command on both the primary and backup system:
STATUS *, PROG RDF-software-loc.*, STOP