I setup a file as a type 19, so I could check it's
items from unix quickly. Not remembering this when
it went production....

I just installed rdiff-backup on our system to make
a mirrored and incremental backups of our system.
http://rdiff-backup.stanford.edu/

When I first attempted to run it, it took almost 5
hours to backup one file.

The problem was it had 180,000 records in it, which
meant (since it's now a type 19), it was a directory
with 180,000 files in it, which posed a minor issue.

Since our system never SELECTed or SORTed the file,
we never saw a slowdown, but man...

Tracking the file down was fairly easy however.

what I did was for each main account.

I did (unix):

du --max-depth=1 -b /accountlocation | tee /tmp/du_output

What this did was give you a byte count of each directory
and tee out the results to a file. You need to watch the
information as it's displayed, when it stops for more than
say 15-20 seconds (or you figure somethings up), write down
the last filename displayed. Do this for each pause. When
completed, vi (or your favorite viewer) the du_output file.
search for the filenames you wrote down, the ones that follow
are the problem childs. If you can capture your screen output
then you can skip the tee.

Once I converted my Type 19 back to a Type 30, it only took
10 minutes to copy. A lot better than 5 hours. Took a while
to convert the file.

I created a new file, copied all the items, deleted the old file
renamed the new file to the old file.
-------
u2-users mailing list
[EMAIL PROTECTED]
To unsubscribe please visit http://listserver.u2ug.org/

Reply via email to