HP 3000 Manuals HP 3000 Manuals
Handling Secondary System Failures
Secondary system failures can be minor or major in the same way that
primary system failures were.
Handling Secondary System Minor Failures
If it is a minor failure when the database on the secondary system is
known to be intact.
1. Run the Manager Program (MP) on SYSTEM A:
:RUN SILMP.PUB.SYS
2. Re-activate the DS lines:
>USE dslines
Silhouette informs you whether it was successful in recovering
from the failure. If it was not able to recover, follow the next
procedure.
NOTE There are no procedures to determine minor failures without
enabling logging (or ROLLBACK). Therefore, secondary system minor
failures are best treated as major failures.
Handling Secondary System Major Failures
A major failure means that the secondary database is not intact.
1. Restore the latest SYSTEM B backup including the databases,
logfile and SAVE file onto SYSTEM B.
a. If the backup was made at the beginning of a log cycle then
purge or rename the SAVE file.
b. If the backup was made during a log cycle or if you using
automatic CHANGELOG, then preserve the SAVE file.
2. On the primary system, issue one or more >USE commands to
re-activate the DS lines and Silhouette to the secondary system.
Handling Failures on Secondary Systems with Disc Caching Enabled
If your secondary system has MPE Disc Caching enabled, a possibility
exists that the Silhouette software will restart successfully even though
the primary system and secondary system databases are no longer
synchronized. A secondary system failure can be minor or major, just as
primary system failures, with regards to whether or not the database is
intact. If the Recovery Process (RP) was updating the database when the
failure occurred, it is difficult to determine if the database was left
intact.
Using one of the two methods described below will ensure the integrity of
your secondary database as well as the secondary system database backups:
* If your system has Intrinsic Level Recovery (ILR) enabled for the
databases, and BLOCKONWRITE enabled, you are ensured of a write
commitment to disc up to the last intrinsic call before the system
failure. This may result in losing the benefits of performance gains
from writes on the devices for which disc caching was enabled. Use
the MPE :CACHECONTROL BLOCKONWRITE = YES command. Refer to the MPE V
System Operation and Resource Management Reference Manual
(32033-90005) for additional information.
* To ensure that the Recovery Process (RP) was not updating the
database when the system failure occurred, check the last STATUS
display to see if the outstanding records are equal to zero. The
primary and secondary logfile record counts should be equal and no
outstanding open transactions should be indicated at the bottom of
the secondary side of the display.
NOTE Recovery for MPE XL systems is similar to those described for MPE V
systems with disc caching enabled. If you have an MPE XL system,
use the second method described above to ensure the integrity of
the secondary database.