architecture

seancorfield 2019-03-13T00:43:40.004400Z

Are those other JVM processes or something completely separate?

2019-03-13T01:38:31.004700Z

I used https://cr.yp.to/daemontools.html for that in the past

2019-03-13T01:39:46.006Z

what you are describing what any init system does, and daemontools just happens to be a sort of init system that is easy to run separately from the init system of the os

2019-03-13T01:40:10.006400Z

but for example, if you are running on linux you could have systemd do it

mathpunk 2019-03-13T02:39:19.015800Z

lets see. in our monorepo: - some docker services are defined - there's a front end built with an npm script - there's a java api not in our repo, - neo4j - occasionally a neo4j-admin command (if I'm setting testing state)

mathpunk 2019-03-13T02:39:55.016600Z

- a selenium server to stand up

mathpunk 2019-03-13T02:41:18.020400Z

I think that's it. In my dreams I can trigger a test run against different commits--- I've already wrapped JGit a bit, for changing what commit the repo is at

2019-03-13T02:43:10.022500Z

Docker compose could start any dockerized services. If you tag your images, you could launch them at those points.

mathpunk 2019-03-13T02:57:39.031400Z

:thinking_face: The services are changing rarely enough that I probably should solve that last. It's the front end and api serving it that change the fastest. Probably that's the slice to bite off

mathpunk 2019-03-13T02:58:49.033100Z

They're the parts I rebuilt a couple three times today

mathpunk 2019-03-13T03:43:43.038900Z

Afterthought: I wonder if those containers, they're full of Java services. Maybe I should start them a different way than the other devs, and learn to e.g. simulate service outages

mathpunk 2019-03-13T15:17:03.047500Z

I think systemd running the selenium server is a good call... That should be running 99% of the time that I'm at work

mathpunk 2019-03-13T15:28:16.048500Z

There's probably two things going on: - being tired of my dumb ass forgetting to, e.g., start the typescript compiler - the more legit work of, sometimes I want to see if a subset of tests that failed will also fail against a different SHA of the codebase

dpkp 2019-03-13T15:49:51.050400Z

make / Makefile would be another possibility to encode dependencies between your services