RDF System Management Manual for J-series and H-series RVUs (RDF Update 13)
1. Determine which disks (the local disk on the primary system or the remote mirror on the standby
system) for all audit trails in the RDF configuration received the most audit records. The example
that follows shows how to do so for the MAT. If your RDF configuration includes one or more
auxiliary audit trails, you must do the same for each auxiliary audit trail.
NOTE: if you had CommitHoldMode configured ON at the time the primary system failed,
then you can omit this step because the mirrors on the remote system will either have more
data or the same data as the local mirrors connected to the primary system. This first step is
only useful if you had turned off CommitHoldMode before the primary system failed.
On the ZLT standby system, use SNOOP READAUDIT to read the final file in the MAT,
positioning at EOF and reading in reverse order for one record. This is sample output from
READAUDIT with the MAT position in bold:
* SEQNO = 8, RBA = 107628804, RBN = 26276 *
AC^RECORD^LENGTH=108, AC_VERSION=7, VERSION_FLAGS=000000 000000, PRIMARY^CPU=0
AUDITING^PROCESS=TMP , VSN=000000 000000 000004 077334
TRANSID=000000 000000 000000 000000, ACTTX=0, TYPE=1033 (DATAVOL STATE)
CREATING^SYSTEM=190, VOLNAME=$DATA13, STATE=8, STATE^TEXT=STARTED
On the former primary system, the last file in the MAT might have been left in the crashopen
state. You can determine that by issuing this command:
$system system 3> fileinfo $*.ztmfat.*
$audit.ztmfat
CODE EOF LAST MODIFIED OWNER RWEP PExt SExt
aa000001 134 125825024 01feb2005 10:15 255,255 gggg 3840 3840
aa000002 134 125829120 01feb2005 10:20 255,255 gggg 3840 3840
aa000003 134 125829120 01feb2005 10:24 255,255 gggg 3840 3840
aa000004 134 125808640 01feb2005 10:31 255,255 gggg 3840 3840
aa000005 134 125829120 01feb2005 10:38 255,255 gggg 3840 3840
aa000006 134 125829120 01feb2005 10:45 255,255 gggg 3840 3840
aa000007 134 125829120 01feb2005 10:54 255,255 gggg 3840 3840
aa000008 134 125829120 01feb2005 11:04 255,255 gggg 3840 3840
aa000009 134 125829120 01feb2005 11:14 255,255 gggg 3840 3840
aa000008 ? 134 107630592 01feb2005 11:04 255,255 gggg 3840 3840
The file marked with the question mark must be fixed. Use the SNOOP FIXUPEOF command
to reset the crashopen flag. Then use SNOOP READAUDIT to read the final record. You cannot
use the MERGE option when specifying the name of the audit trail file. Because the TMF
product is not started, attempting to use the MERGE option results in an error. Using the
example of the MAT, specify the MAT volume and subvolume when SNOOP issues this prompt:
Audit trail name or 'MERGE' (MERGE): $AUDIT.ZTMFAT.AA
Compare the MAT position of the two records to determine which disk has the most audit
records.
2. Recover the database on your former primary system. How you do this depends upon whether
local disks or remote mirrors received the most audit records (which you determined in the
preceding step).
CommitHoldMode ON
If all of the remote mirrors (MAT and all auxiliary audit trails) have more or the same number
of audit records as the local disks (this typically happens if CommitHoldMode was configured
and enabled on the primary system when the outage occurred):
a. Issue SCF STOP $audit-vol on the former primary system (this stops the local disk).
b. Issue SCF STOP $audit-vol on the ZLT standby system (this stops the remote mirror on
the ZLT standby system).
c. Issue SCF START $audit-vol -M (this starts only the remote mirror).
d. Once the remote mirror is started, issue SCF START $audit-vol (which causes the revive
from -M to -P)
Recovering the Primary System After an RDF ZLT Takeover 337