HP NonStop RDF, Руководство по управлению

Страница: 140 / 478

Transactions that must be undone during this undo pass are stored in the ZTXUNDO file in your
Master Image Trail subvolume. You can use the READLIST utility to see what transactions were
undone by this Local Undo operation.
Phase Two Undo Pass
This is also known as File Undo. If one or more volumes failed on the primary system and a
transaction aborted, the TMF Backout process will backout the transaction on the volumes that
are still up, but it will be unable to backout the audit on the volumes that are down. If the downed
volumes come back online, the TMF Volume Recovery process backs out the audit that the
Backout process could not back out. If, however, the primary system failed before Volume
Recovery had enabled the downed volumes, then, if you execute the RDF Takeover operation
on the backup system, the updaters execute an undo pass that will undo the audit that Volume
Recovery would have undone on the primary system if it had been able to.
Transactions that must be undone during this undo pass are stored in the ZFILUNDO file in the
Master Image Trail subvolume. You can use the READLIST utility to see what transactions were
undone by this File Undo operation.
Phase Three Undo Pass
This is also known as Network Undo. If you are running in an RDF network and you lose one
or more primary systems, you must do a takeover on all backup systems in your RDF network.
For a complete description of the takeover operation in an RDF network, see
“RDF Takeovers
Within a Network Environment” (page 298) .
Transactions that must be undone during this undo pass are stored in the ZNETUNDO file in
the Master Image Trail subvolume.
Issuing the TAKEOVER Command
Before you issue a TAKEOVER command on your backup system, you need to start an RDFCOM
session for the correct control subvolume. For example, if you were running RDF from \Boston
to \London and the control subvolume was BOSTON, then you would enter the following
RDFCOM command to the TACL prompt on \LONDON:
> RDFCOM BOSTON
The Takeover command has the ! option (see the syntax for this command in Chapter 8 ). If you
do not include the ! sign, then RDFCOM tries to see if the monitor and extractor are still running
on the primary system. If it can access the monitor or extractor because the primary system is
still running, then RDFCOM aborts the command immediately. If you include the ! option, then
RDFCOM does not try to reach the monitor and extractor on the primary system. The ! option
also determines whether or not you are prompted to confirm your intention of performing the
Takeover operation, a topic discussed a little further below.
It is highly recommended that you do not use the ! option because it prevents the Takeover
command from getting started if the primary system is still running. If this is the case and if you
execute the Takeover operation anyway, the result is that the primary and backup databases
become out of synchronization.
140 Critical Operations, Special Situations, and Error Conditions