News for hackers. Stuff that matters.

lipsyclouds

Re:Bob CD on Monday November 14, 2022 @05:40PM (#33602388)
hello! i am taking this as valuable feedback and would improve the website more! like you suspected, this is more of a project ive been building to solve problems ive been facing and didnt really do much publicity hence the website could definitely use some more love!
Re:Bob CD on Monday November 14, 2022 @05:24PM (#33602164)
hello, the author here! all of these points have been taken as valuable feedback and thanks a lot for this! what @capableweb said is pretty much the goal of bob, primarily originating from my experiences with various tooling along my career and pretty much built in my free time. would love if more people have a similar perspectives and would love to see it in a more formal way that just a way to solve my problems :) and of course happy to collaborate and PRs most welcome! :D
Re:Bob CD on Monday November 14, 2022 @05:19PM (#33602109)
hello! the author here! the primary reason for going with this architecture is to make it scale well with off the shelf production ready tooling. RabbitMQ was chosen to provide a scalable durable queue for backpressure support and load balancing across runners, error handling etc. it adds onto the infra weight a bit but seems to be a simpler choice to program against and use libraries. The reasons for XTDB are mainly 2: it provides a temporal immutable store, hence the whole state of the CI system can immutably evolve like a git repo hence the whole thing can be inspected at a previous time. Bob treats time as a first class thing and not just build logs but you should be able to see the whole cluster at a past time and debug it fully as all of the state is the DB. The second reason is like @capableweb said, it decouples storage from compute so that we can have a pluggable storage and not a central point of compute.
Re:Bob CD on Monday November 14, 2022 @05:13PM (#33602027)
hello! the author here! this has been pretty much made with the intention of me trying to address some of the shortcomings i experienced in the kind of tooling ive used over the years as an infra/sre person. never made it with the intention of making it general enough tool for everyone but more as a manifestation of what i think tools could've been like. it draws a lot of inspiration from concourse but improves on aspects like decoupling the way one can describe things for instance in YAML like concourse, no central controller in the cluster and things like being designed for being cloud native upfront by having rootless unprivileged runners etc. the way i see it concourse is a production ready, general version of this and other people like the idea of bob, it could be too by evolving more on the ideas of concourse.