I setup a global poliicy for compression in the Kopia UI. I want to make sure all my newly setup snapshots uses the best compression.,
The defined portion of the compression pane is empty, it does not look like it is inheriting from the global policy but the effective part shows the compression I set up. Is that how it works?
‘effective’ is what will actually happen, defined is a local override.
Setting policy at a global level applies to all snapshots, unless more specific “defined” is present at a user, host or a directory level.
Thanks for the quick reply.
So this is how my Policies page look like (not showing my snapshots for privacy). Does this look correct?
As you start creating new snapshot in the UI you will see the actual/effective values recomputed live every time you change the directory, this should confirm the values.
BTW. I recommend against
zstd-best-compression - it has abnormally high memory usage compared to compression beenfits and is causing trouble for folks.
Thanks, can I change the compression type after the initial snapshots taken? You are right about the mem usage. I am taking snapshots of large folders (many hundreds GBs now and Kopia is using 33GB, half of my physical memory.
Today changing compression after the snapshot is not possible, but thould be doable in the future. PRs welcome.
It goes up to 100gb+ of memory use with a snapshot. that seems insane, abnormal does not seem enough to describe this. Are you sure this is not memory leak?
What other good option do you recommend for compression?
I have tested it with a snapshot repo wth 7.2GB and 35k files. Kopia memory usage went over 105GB (private bytes)on Windows 11, this seems really insane to me.
Also once the snapshot is done Kopia is not releasing that memory quickly, it seems to take a long time like 5-10min per 10gb.
Is anyone else experiencing this kind of memory usage?
I usually settle for one of the s2 versions. Getting the few percents of performance is not at all as important as the speed is (in my personal view!) and for really simple data, all compressors do a fine job anyhow, so chasing <5% in size for 10x the cpu-time required to find a few extra bytes to save doesn’t feel worth it.