Fork me on GitHub

Hi @michaeldrogalis, just wanted to know if there is any news on the onyx tag issue that was mentioned some time ago?


@atwrdik Not yet - sorry that one fell off my list. Can you refresh my memory: there was a patch for that area of the scheduler that we were going to try on your case -- did we see what happened there?


@michaeldrogalis in onyx version 0.12.2 a submitted job would not start if tags were used. The error message was something like: "Job ID X has been submitted with tenancy ID Y, but received no virtual peers to start its execution." When switching to onyx version 0.12.3 the submitted job actually started but tasks were assigned to peers not capable of performing the desired action. It seemed as if the tag information was plainly ignored.


Back then I set up a repository containing a minimal reproducer for onyx version 0.12.2:


@atwrdik Right, thanks. I'll try to get a look at it later today. Quite busy over here lately, but we'll get it down.


@michaeldrogalis thx in advance. Please don't put too much pressure on you. Take your time and just gimme a sign if there is any news on the topic šŸ™‚


@atwrdik Not a problem, it's gotta get done at some point. šŸ™‚ Ill let you know.


I've been bouncing around the idea of doing a distributed implementation of the Rete algorithm using Onyx. Overall, it seems like a great fit. The one thing I'm puzzling over is whether it's possible to have segments generated by tasks fed back into the job. These internally-generated segments must be processed before moving forward to handle any new external segments. Is this even possible within a single job? I'm sure I could find a way to make it work with some sort of external component, but curious if it's possible entirely within Onyx.


@dave.dixon Iā€™m imagining a flow condition that allows some circular workflow between tasks. The flow condition would keep from infinite loops and check to see whether it was newly generated data that should be spent back up the workflow for processing or not.


Onyx doesn't support cyclic flows at the moment - we think that's best achieved by involving an outside data source.