HP 3000 Manuals HP 3000 Manuals
Effects of Operating System Priority
The master and slave resynchronization applications run from a job script
should have a high enough priority to make sure they have sufficient CPU
time. The application of transactions on the slave should keep up with
the committing of user transactions on the master.
If the resynchronization applications do not get enough CPU time on the
target system, they will probably start to fall behind the applications
that are committing user transactions against the master DBEnvironment.
If they fall too far behind, a hard resynchronization may become
necessary. This situation can usually be avoided by ensuring that the
resynchronization applications are run at the same or higher priority
than the applications that are committing the user transactions.
Using SQLAudit to Check if Slave is Keeping Up
[REV BEG]
Periodically check how far the slave is falling behind the master. To
use the technique below, you must be sure that audit logging has been
enabled on the slave by specifying the AUDIT LOG parameter in the most
recent START DBE NEWLOG statement. (Use the SQLUtil SHOWDBE command to
verify that audit logging has been enabled on the slave. If it has not,
use the SQLUtil START DBE NEWLOG command to enable audit logging.) This
will cause audit log records to be generated on the slave each time a
transaction from the master is committed on the slave. These records can
then be audited to determine how well the slave is keeping up with the
master.
To check how well the slave is keeping up with the master, use the
following steps:
1. Use the SQLAudit GET AUDITPOINT command to get an audit point.
: SQLAUDIT
SQLAudit >> SET DBEN PARTSDBE
SQLAudit >> GET AUDITPOINT
Audit Point File >> TTEST1
Lock Log for Audit Point (n/y) >> N
Display Audit Point Information (n/y) >> N
2. WAIT for one minute. (If you have very long transactions you may
want to wait longer.)
3. Use the SQLAudit AUDIT command to audit the transactions that have
been processed by ALLBASE/Replicate since the last audit point was
established.
SQLAudit >> AUDIT
Beginning Audit Point File >> TTEST1
Ending Audit Point File >> Return
Using Current Audit Point Information (obtained from current DBEnvironment).
Results File >> TTESTR1
Do you wish to specify Partition Numbers (n/y) >> N
Generating Audit Results ...
Records Audited: 3454 Records Generated: 3454
Finished Generating Results.
SQLAudit >> EXIT
:
4. View the results file using your favorite editor. Look for
records similar to the ones below:
COMMIT User: USER1 Audit Name: madbe1 Time: 1994-11-07 19:29:46.358
Original Time: 1994-11-07 19:00:01.206
COMMIT User: USER1 Audit Name: madbe1 Time: 1994-11-07 19:30:26.909
Original Time: 1994-11-07 19:26:58.936
The value specified for "Time" is the time the transaction
committed on the slave. The value specified for "Original Time"
is the time the transaction committed on the master. You can tell
how far the slave is behind the master by noting the time
difference between the two values.
[REV END]