Mark,
I am not sure I understand how your solution
of backing up the log less often will help.
The same amount of data will need to be backed up
whether you queue them up or not.
Each log backup, regardless of whether it is
10 minutes or 10 hours apart is represented
by a unique backup object for each an every
".TXN" file. They are not "grouped" together
into a single object on the TSM Server.
This is because each log backup mat need to be
restored at any given time.
... or are you after some behavior that will
keep all of these log backups together in
a better fashion if they are backed up at nearly
the same time? For example... if you are sending the
logs directly to tape... then this might produce
the behavior that you are looking for.
I couldn't tell from how you described this.
You mentioned "DLT" but didn't say backup
directly to tape... an important point.
Brian's technique (of hourly log backups)
is a popular one because it also covers
the situation of a disk failure. That is if the
Domino Server transaction log disk were to crash
and there was no "hardware" redundancy, the
hourly backing up of logs would reduce the exposure
to a minimum loss of data of less than an hour.
In my last append to this subject, I described one
setup that results in any database being restored
and logs applied to that database in 8 minutes or less.
Your mileage may vary depending on your TSM Server platform,
network speed, etc... but it is just one example.
Thanks,
Del
----------------------------------------------------
Del Hoobler
IBM Corporation
hoobler AT us.ibm DOT com
- Never cut what can be untied.
- Commit yourself to constant improvement.
=========================================================
> ...and herein lies a future preventative. Is it *really* necessary to
> backup log files every hour?
>
> Suppose you do hourly log backups after a weekly full backup on Sunday
> at 0100. You have data loss at 1500 on Friday, and start a restore. You
> will have to restore the last full backup, and then you have to restore
> *134* separate log backups. If you're using, say, DLT tape, and not
> collocating, you may get to run 135 separate mounts, spin-forwards,
> restores, log replays, rewinds, and dismounts for each file you're
> restoring. Each transaction requires a minimum of 2 minutes (if there's
> little or no spin-forward/rewind) plus, say, 1 minute of log playback. 3
> minutes/transaction times 135 transactions equals 405 minutes, or more
> than 6 hours per file. Minimum. If you have to run to the end of a tape
> for a log, add 6 to 8 minutes per transaction. Each file restore *could*
> run as much as 870 minutes. The real world is, of course, somewhere
> inbetween, but an average of 10 hours per file isn't pretty, no matter
> how you slice it.
>
> Such is the curse of (necessarily) single-threaded restores. Do yourself
> a favor. Cut your log backups to 2 or 3 a day. Your Domino group won't
> like it, but do the math for them. They'll like it better then.
|
