Could vzdump be modified to allow for a different compression binary, other than gzip? Ideally, allow for any binary (as long as stdin is supported) and options in the vzdump config file. I'd really like to make use of some of the more modern / advanced compression schemes now available, specifically some of the multithreaded ones. The gzip process pegs one of the cores on our extremely fast Core i7 right now, getting rid of that bottleneck would be great!
vzdump, different compression schemes
- Thread starter obrienmd
- Start date
Specifically, pigz support would be awesome - it's even in the Debian Sid repositories.
OK, tried this the simple way by installing pigz from debian repositories (aptitude did not work, had to download sid 64-bit dpkg, probably don't have the right repository turned on or something). After install, I renamed /bin/gzip to /bin/gzip.old, and made a symbolic link from /bin/gzip to the pigz binary.
This fails because vzdump apparently calls gzip with the -R option, which (because it is not yet supported in pigz) errors out pigz.
So, couple of questions:
1) Is there an non-obvious reason, other than rsyncing is more efficient (which we do not use in our backup transfer routines anyway), that the -R option is used for gzip in vzdump?
2) If not, how can I remove this option? I browsed through the vzdump script, but did not see gzip called and don't know perl, so gave up.
This fails because vzdump apparently calls gzip with the -R option, which (because it is not yet supported in pigz) errors out pigz.
So, couple of questions:
1) Is there an non-obvious reason, other than rsyncing is more efficient (which we do not use in our backup transfer routines anyway), that the -R option is used for gzip in vzdump?
2) If not, how can I remove this option? I browsed through the vzdump script, but did not see gzip called and don't know perl, so gave up.
it is used because many people use rsync.
If you don't know perl, you should learn it - or what is the question?
Do you really want the backup to use all available cores for compression? Most people do not want that.
I would like to be able to specify, ideally (as pigz allows) - but yes, all available cores would be great. It would be unlikely to use 100% of all cores, of course.
Also, WRT your last question in the previous response, I'm asking how to modify the command line where vzdump calls gzip, so I can remove the -R for now. Good suggestion, to learn PERL, though
Why. If you allow to use all cores it will use all cores. So it is quite likely that the backup will utilize all cores.
Also, WRT your last question in the previous response, I'm asking how to modify the command line where vzdump calls gzip, so I can remove the -R for now. Good suggestion, to learn PERL, though
Yes, the suggestion was that you should try to find that yourself ;-)
Anyway, vzdump does not use '-R' at all, instead it uses '--rsyncable'. Its on line 32 of vzdump:
Code:
# by default we set --rsyncable for gzip
$ENV{GZIP} = "--rsyncable" if !$ENV{GZIP};