On 2018-04-04 06:01, Stefan G. Weichinger wrote:
Am 2018-04-03 um 20:52 schrieb Austin S. Hemmelgarn:
On 2018-04-03 14:25, Stefan G. Weichinger wrote:
Does anyone already use zstd https://en.wikipedia.org/wiki/Zstandard
with amanda?
I will try to define an initial dumptype and play around although I
wonder if the standard behavior leads to any problems.
zstd does not remove the source file after de/compression per default
(only with "--rm") ... but as it is used within a pipe (?) with amanda I
assume that won't hurt.
The "-d" for decompression is there so that should work.
I've been using it for a few months now both at home and at work. It
works just fine as-is and gets pretty good performance.
In both cases though, I actually use a wrapper script. The one for
backups at work just adds `-T2` to the zstd command line as our backup
server has lots of CPU (and CPU time), but the backups are
network-limited. At home, I also bump the compression level as high as
I can without needing special decompression options (so the full command
line at home that the wrapper passes is `-19 --long --zstd=hlog=26 -T2`).
I've done numerous restores from both sets of backups both with and
without the wrapper script (I initially set both up to just use zstd
directly), and it all appears to work just fine.
Would this work as well?
That's essentially what I used initially, and I had no issues with it at
all either backing things up or restoring.
->
define dumptype client-zstd-tar {
global
program "GNUTAR"
comment "custom client compression dumped with tar"
compress client custom
client_custom_compress "/usr/bin/zstd"
}
