GitLab GitLab Group Kickoffs - Enablement:Gitaly, 15 Sep 2020

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: Gitlab 13.5 kickoff - Create:Gitaly

Description

Planning issue - https://gitlab.com/gitlab-org/create-stage/-/issues/12709
Direction page - https://about.gitlab.com/direction/create/gitaly/

A

Hi, this is derek ferguson senior product manager for the getaly group here at gitlab, and today I'm going to be going over what we are planning on working on in the 13.5 release of git lab. So the first thing that we are planning on is the general availability of the giddily cluster, strong consistency.

A

So, with the addition of majority winds voting strategy in 13 4, we will be on course to deliver strong consistency as ga in the 13.5 release, we're just working on cleaning up a few of the lingering issues and it should be available for general availability.

A

The next thing we're working on is allowing giddily clusters to have a variable replication factor, and this means that, while currently each repository in a giddily cluster is replicated across all the nodes within that cluster, this feature will allow users to specify how many times a repository is replicated across those nodes.

A

So this will help with storage as well as horizontally scaling giddily clusters, as well as several other features that we have planned, such as auto rebalancing.

A

So the first thing that we are working on is being able to specify the primary giddily node per repository. Currently, every single repository on a cluster uses the same giddily node as the primary. By changing this to b per repository, it frees us up to act for each repository as a separate entity and allowing multiple nodes to be set as primary nodes. This removes the bottleneck of a single node for every single repository being the primary.

A

The next thing we're working on are some performance improvements in moving a lot of our code from ruby to go so with this, we plan on getting a lot of improvement in the performance of our rpc calls, as well as several other things that we believe that this will give us some performance boost on and finally, a long going project that we have right now is improving the native get storage of large binary files and specifically we're still working on the plain http driver for promisers.

A

So this will allow us to use an http protocol for large binary objects that are better suited to a protocol that can support resuming as streaming objects rather than the git protocol. So this is an ongoing project. It's a long running project, but we're still working on it and those are the main things that we have planned for uh giddily in 13.5.

A

Thank you very much.