Thanks to everyone who participated, and apologies for the delay in getting
this posted.

video recording: https://youtu.be/6zb6AxL26d8

notes (thanks Serapheim):

   -

   Corrective Recv PR <https://github.com/openzfs/zfs/pull/9372> (Alek)
   -

      The functionality introduced here is for being able to “heal” a
      corrupted dataset from a send file (instead of deleting and
recreating it).
      An interesting details is that this feature works regardless of the
      differences in the configuration of the target machine and the
machine that
      generated the test file (e.g. compression or other properties).
      -

      Current status is that most of the tests are passing except a couple
      from the redacted send-receive tests in FreeBSD which cause a panic.
      -

         John Kennedy and Paul Dagnelie can help with this
         -

      Besides fixing the panic - the PR needs reviews.
      -

   Add Blake3 Checksum (PR <https://github.com/openzfs/zfs/pull/11897>)
   (Tino)
   -

      Currently experiencing problems with supporting this on FreeBSD. The
      complications seem specific to the code organization and getting
the right
      directives in autoconf/automake and make to work.
      -

      Besides just adding this new checksump algorithm, the community would
      like to see how this algorithm compares to existing ones, running on a
      sufficiently modern CPU under different blocksizes.
      -

   Should I add sha_ni (x86) and maybe also aarch64 sha2 hardware support?
   -

      There is probably some low-hanging fruit for performance
      improvements, especially for our choice of CPU instructions.
      -

      Some support for this in FreeBSD has started being worked on:
      https://github.com/freebsd/freebsd-src/compare/main...allanjude:zfssha
      -

   Interface for /proc/spl/kstat/zfs/hashing_bench ? echo blake3-sse2 >
   hashing ..
   -

   ZFS and Linux Containers (LXD) (Allan Jude)
   -

      The idea here is jailing a dataset (group of datasets) to a user
      namespace in LXD.
      -

         In terms of code, this would mean at least implementing the
         IS_GLOBAL_ZONE() macro that currently in Linux always returns true.
         -

      Do people have interesting use/test cases?
      -

         If you do, please do reach out to Allan.
         -

      What makes sense for extending the ZTS to cover containers?
      -

         Besides the actual ZTS-specific additions, we’ll need to look at
         what we can reasonably add to our automated testing bots in
Github to run
         those tests.
         -

   Cross-platform compatible xattr support (Alexander Motin - amotin)
   -

      Current status: There is progress but we’d love it if more people are
      able to look at the code and/or bring alternative solutions/ideas to the
      discussion.
      -

   ZFS on Object Storage
   -

      Currently an experiment within Delphix
      -

      The idea is to run ZFS on top of an object store like S3 or MinIO.
      -

      The focus is on performance for small record sizes.
      -

      There are currently two brand-new components being introduced:
      -

         A userland agent that facilitates the interaction between ZFS and
         the object store
         -

         A new block-based local cache introduced as a vdev type to improve
         the performance of the overall design of this project.
         -

            The L2ARC was briefly considered but it doesn’t satisfy the
            requirements for the current use-case as its memory
requirements are high.
            -

      More technical details on the upcoming OpenZFS Summit.
      -

      Until then, while the technical requirements are still being designed
      we’d love any feedback on the administrative/user interface of
this project.
      -

         Currently we just provide a couple of user properties just to test
         the design (e.g. a property for the s3 URL, credentials file, etc..).
         Eventually if we are to start supporting other protocols
we’ll need to have
         a user interface and workflow that works for all of them.


On Mon, Apr 26, 2021 at 11:49 AM Matthew Ahrens <mahr...@delphix.com> wrote:

>
> The next OpenZFS Leadership meeting will be held tomorrow, April 27, 1pm
> Pacific time.  Topics on the agenda include:
>
>> Userland zfs (PR <https://github.com/openzfs/zfs/pull/11766>)
>>
> Everyone is welcome to attend and participate, and we will try to keep the
> meeting on agenda and on time.  The meetings will be held online via Zoom,
> and recorded and posted to the website and YouTube after the meeting.
>
> For more information and details on how to attend, as well as notes and
> video from the previous meeting, please see the agenda document:
>
>
> https://docs.google.com/document/d/1w2jv2XVYFmBVvG1EGf-9A5HBVsjAYoLIFZAnWHhV-BM/edit
>
> --matt
>

------------------------------------------
openzfs: openzfs-developer
Permalink: 
https://openzfs.topicbox.com/groups/developer/Td7b3a91ecbeec509-M69f8c634d2c7adb5b176c800
Delivery options: https://openzfs.topicbox.com/groups/developer/subscription

Reply via email to