| .0> As secondary archiving is just used to empty the Archive database and the
.0> store directories I thought of a dirty workaround.
.0> i.e. use SQL to delete the records from the Archive database (delete from
.0> TLF_TABLE where credate_q <= 'date time'; the same also applies to the
.0> other tables THF_TABLE, CHF_TABLE and CLF_TABLE).
.0> and also delete everything from the store directories
.0> (delete/before=date time)
Experience has shown that this method, as described, will inevitably cause
problems, e.g., orphan files, too may files deleted, orphan history records,
too many history rcds deleted. A modified version of this method,
selecting tlf/clf by date/time, but deleteing tlf/thf/clf/chf/store_files by
selected ids, does work much faster, and has less "weird failures/recoveries"
known to secondary archive processes.
.0> Is there a way to increase the backup speed. ?
- Insure queue lengths, for physical and logical disk devices, do not
exceed 0.5 (there are many, many methods for achieving this).
- Insure there is always >25% free bytes in each storage area, for both
audit and archive databases (due to poor design and index placement).
- No edi version listed, however if <v2.0, optimized arch_data .dat's thru
optimized fdls help tremendously.
- INTERCHANGE ARCHIVE user's UAF account with large biolm, diolm, wsextent
(not exceeding sysgen's wsmax), helps.
I can talk tuning/timing-n-sizing all day, but it's always a config-by-config
decision.
...Mike
|
| Hello Mike,
thanks for your answer.
It's V2.1C of DEC/EDI.
Imagine that the system is tuned as stated by you, how long would it take to
archive 40,000 documents using INTER ARCHIVE ?
.1> Experience has shown that this method, as described, will inevitably cause
.1> problems, e.g., orphan files, too may files deleted, orphan history records,
.1> too many history rcds deleted. A modified version of this method,
.1> selecting tlf/clf by date/time, but deleteing tlf/thf/clf/chf/store_files by
.1> selected ids, does work much faster, and has less "weird failures/recoveries"
.1> known to secondary archive processes.
Do you have any examples ?
Thanks,
Martin
|