Where communities thrive


  • Join over 1.5M+ people
  • Join over 100K+ communities
  • Free without limits
  • Create your own community
People
Activity
    andrewchambers
    @andrewchambers
    and work out a diff to fetch
    so it should be possible to resume a slow fetch if something goes wrong
    sogaiu
    @sogaiu:matrix.org
    [m]
    that sounds nice!
    andrewchambers
    @andrewchambers
    Hmm, I found some simplifications that make the directory storage backend go a fair bit faster.
    actually for my external drive 3x faster
    sogaiu
    @sogaiu
    oh nice!
    andrewchambers
    @andrewchambers

    Hey everyone, just pushed a new release - bupstash 0.8.0.

    Main new feature is the addition of bupstash diff.

    Change log here: https://github.com/andrewchambers/bupstash/releases/tag/v0.8.0

    bket
    @bket
    andrewchambers: thanks for the update! #123 decreases backup time significantly
    Question: i have pledged/unveiled (OpenBSD only) bupstash using bket/bupstash@3a243c0 ...
    ... and was wondering if it the time is right to open a PR
    s/if it/if
    andrewchambers
    @andrewchambers
    sure, anytime time is fine by me.
    i might end up doing a 0.8.1 release as i tidy up the documentation
    andrewchambers
    @andrewchambers
    I might modify the init code to make your change easier
    currently it looks like i create repo.tmp
    then move it into position
    Tscho
    @tschuxxi
    Hello. I installed bupstash on a couple of machines now and have to say I like it pretty much so far. A couple of questions came up: Is there a way to set the timestamp? Since I backup a database, a config file and a folder I would like to see them backed up at the same time.
    Is there (yet) a way to tell the rm to keep the latest 24 hourly, 7 weekly, 30 daily and 6 monthly backups?
    andrewchambers
    @andrewchambers
    @tschuxxi not for the first question, I'm not sure how i would make that work exactly, though it might be possible.
    for the second question, i think the best way will be for me to make the --format=jsonl format stable
    and for a custom script to do it
    however, it might be common enough we can build it in
    I will need to think a bit about how the command line for such a command might look
    I will try to make a guide for how to do this when it is added as i think it will be a common request.
    andrewchambers
    @andrewchambers
    I've created this ticket to track this request
    andrewchambers
    @andrewchambers
    @tschuxxi i thought more about letting the user set the timestamp, its a bit more complicated than i thought because the timestamp has some meaning for security
    i am wondering if i should swap from client side timestamp to server side timestamp
    in which case allowing the user to set the timestamp would not be possible at all
    as if your client machine was hacked at a certain date you can determine
    then you want to only use backups from before that time
    so using server side timestamps makes more sense in that case if you have setup a security boundary.
    bket
    @bket
    andrewchambers: I had another look at the pledg-bits in PR #132
    i really do not like unveil'ing parentdir of repodir
    instead we could unveil repodir.bupstash-repo-init-tmp in repository.rs
    actually we need to unveil it 2 times: one for testing/creating repodir.bupstash-repo-init-tmp, and one for filling repodir.bupstash-repo-init-tmp
    andrewchambers
    @andrewchambers
    @bket yeah, i think i can maybe make init just work with a single directory
    to avoid the tmp thing, sorry i've been a bit slow, i have had a few things to work on
    It has not slipped my mind though
    andrewchambers
    @andrewchambers
    Been working on adding more parallelism to 'bupstash put'
    i think im most of the way there
    the idea being 'bupstash put' can form a parallel pipeline where it can do
    read+hash -> compress -> encrypt -> send all in parallel, should be pretty fast
    potentially 3x speed improvement if there is no disk involved and you have 3 cores
    andrewchambers
    @andrewchambers
    darn, looks like i was too optimistic
    sogaiu
    @sogaiu
    it wasn't as fast as you thought?
    andrewchambers
    @andrewchambers
    @sogaiu yeah, it got about 10 percent faster for a huge backup
    when i was encrypting and compressing in parallel
    even when there was no disk io
    which sort of surprised me