Absurd In Production
12 points by facundoolano
12 points by facundoolano
Looks promising.
I've been using DBOS at work for the last few months, which is another DB-backed durable workflow system that aims to be simpler than Linear, but it's freemium, not fully open-source, with weird limitations:
it doesn't have partitioning
Absurd currently does not do partitions yet. It's a thing I definitely want to support, but I have not found the optimal design yet. I have a few ideas but none are explored. I'm glad for any input in that regard (that does not put it down on a path of dependency on proprietary database extensions).
Yes, I read your post and your GitHub issue https://github.com/earendil-works/absurd/issues/4.
I was referring to DBOS (the open-source version) not having partitions and other shortcomings, which is why I am interested in Absurd. I actually forked DBOS to add partitioning (https://github.com/Dirac-Software/dbos-transact-py) and rename scheduled operations to make them partitioning-friendly. It uses a hacky maintenance stored procedure to be executed from cron (not pg_cron).
As you point out, expires_at is not necessarily just in the current or next week, so this naive approach of pre-provisioning the next N weeks' partitions wouldn't work in the general case. I would expect something based on pg_partman could work, but it is async would require some cooperation with the Absurd core. I would venture to add a system like Absurd would benefit from an admin command to quiesce the system for maintenance windows.
I wonder how many many of the intervening changes since the original vibecoded implementation were also vibecoded?
I don't think anything in Absurd was "vibecoded" (as in: give in to the machine, don't look at the code) but it has been built almost entirely with the help of coding agents as this has become the predominant way in which I write code.
I'm not a huge fan of writing SQL and having the machine in doing that for me has been a very pleasant experience. The SDKs have quite a bit of hand written code in them.
I'm not a huge fan of writing SQL and having the machine in doing that for me has been a very pleasant experience.
🤝