TMF Planning and Configuration Guide (H06.05+)

Fallback Considerations
HP NonStop TMF Planning and Configuration Guide—540136-002
A-3
Fallback From Phase C
Because prior RVUs of TMF do not support multiple file sizes for a given audit trail,
some of the old algorithms do not lend themselves to handling multiple file sizes. In the
best case, values for overflow locations and abort transaction locations would be
incorrect (potentially hurting system performance). In the worst case, if the file size was
decreased for some reason prior to fallback, the backout algorithm could fail (resulting
in a TMP crash).
Fallback From Phase C
If you must fall back to a format 1 RVU after switching to format 2, you should
transition back to format 1, eliminate all format 2 audit-trail files, and then get new
dumps so that no format 2 audit-trail files are needed after fallback.
However, if fallback must happen quickly and format 2 audit-trail files exist, you must
delete TMF (because format 2 audit-trail files are not compatible with format 1 RVUs of
TMF). Note that if you issue DELETE TMF, the size of the audit-trail files becomes
important.
File Size Greater Than 4 GB
If the file-size attribute of one or more audit trails is greater than 4 GB, you cannot
delete TMF until you purge all greater-than-4GB audit-trail files by using SNOOP
PURGEAUDIT commands. After every audit-trail file greater than 4GB has been purged,
you can issue the DELETE TMF command.
File Size Less Than 4GB
If the file size of all audit-trail files is less than 4 GB (whether altered or not), DELETE
TMF should work without having to use SNOOP.
Note. If you decrease the file-size attribute of an audit trail then immediately need to fall back
to a format 1 RVU in a non-clean state where transactions require backout, the START TMF
command after fallback might attempt to access files for backout that have EOFs greater than
the currently configured file size. In this case, the TMP may crash while doing some
calculations to determine which transactions to back out for a given run of backout. To avoid
this situation, before falling back alter the file-size attribute back to the original (greater) value
for any trails that are in transition where the file size was decreased. You can do this while
stopped immediately before falling back. This action does not change any files, but does
change the TMF configured file size in the AUDINFO file for the given trail(s), and could
prevent a serious problem after falling back if TMF accesses files that have EOFs greater than
the configured file size for that trail.
Note. You might avoid a DELETE TMF if fallback is necessary after altering the TMF audit-trail
format. When considering altering the audit-trail format, first read the disaster recovery support
note. If you follow the steps indicated by the support note for disaster recovery, you might avoid
the DELETE TMF and instead recover the database to a specific point in time, perhaps
immediately prior to altering the audit-trail format.