apache log backups

Darxus at ChaosReigns.com Darxus at ChaosReigns.com
Tue Jan 9 21:02:14 GMT 2007


I'm watching my backup via rsync, throttled to a very low speed.  Looks
like downloading apache logs is taking the longest time (when I'm rsyncing
over an old copy of the same data) because it's not noticing that, for
example, www.chaosreigns.com-access.log.196.gz on the origin is the same
file as www.chaosreigns.com-access.log.186.gz on the destination, so
*everything* is getting unnecessarily retransmitted.

How do people deal with that?

In this case bandwidth is basically all that matters (I'm backing up 7.5gb
of data with --bwlimit=8), time and cpu are not very important.  So I was
wondering if it was possible to tell rsync to pay more attention to
checksums than filenames / paths, so it can notice what the above file got
renamed to and mimic the rename without a retransmit?  Man page is looking
like no, any chance of that getting implemented?  Or is checksumming two
copies of 7.5gb of data not something I actually want to do (p3 550 -> A64
x2 3800)?

The other obvious option would be to use dates when renaming logs instead
of incrementing log numbers.  I'm kind of surprised logrotate doesn't seem
to support that.  Wouldn't be too hard to write a script to do it pretty
cleanly.

Anything I'm missing?

-- 
"For gasoline vapor, the explosive range is from 1.3 to 6.0% vapor
to air...useful against soft targets such as...armored vehicles...and
bunkers." - http://www.fas.org/man/dod-101/sys/dumb/fae.htm
http://www.ChaosReigns.com


More information about the rsync mailing list