Hi guys.
Is it possible - even if not officially covered/supported - to tweak Discourse/container to start/boot faster? (outside of tweaking resources of the host |& container itself)
AI-generated summary
The discussion revolves around speeding up container instantiation for Discourse. lejeczek seeks ways to tweak the container to boot faster, excluding resource tweaks. Lhc_fl suggests using multiple container files and provides an unofficial method for installing plugins without rebuilding the container.
lejeczek clarifies that they want to reduce the downtime associated with restarting the container, not rebuilding it. pfaffman explains that the startup process takes time due to Rails and dependencies, but suggests a two-container solution, separating Rails and Nginx from Postgres and Redis, to achieve zero downtime.
lejeczek inquires about the two-container solution, and pfaffman confirms that it involves running separate containers for the application and database, allowing for faster restarts and zero downtime.
You can try to use multiple container files instead of a single standalone app.yml.
But let’s be honest, almost no one needs to rebuild containers frequently, right?
If you need to install plugins, it is not necessary to rebuild the container. I have unofficial ways to install plugins without rebuilding.
Do you really mean boot faster, that is, start serviing pages sooner after you restart the container? No. It just takes a while to get rails and all of the stuff up and running.
If you mean the downtime associated with rebuilding a container, the the two-container solution is what you’re looking for.
And if that’s not fast enough and you want zero downtime, then you can do stuff with a load balancer and spin up the new container while the old one continues to serve pages and then shuts down the old one after the new one is ready.
I do not (necessarily) mean rebuild.
Say one manages - as I’d prefer - the whole lot with docker
Even a docker restart of container seems to have Discourse do whole long list of things as it gets boot up — yes, that was what I was meaning — can this be tweaked/trimmed, Ruby/Rails stuff? (Why is it that every start all that gets done? Can that not persist?)
Two-container solution — what is that?
Is it two containers with relevant bits not colliding and on the same one host talking (both containers) to the same, external both pgSQL & Redis?
Or is that a zero-downtime setup?
It has to read a whole bunch of code to be ready to serve it. Have you ever booted your phone, a computer, or a modern television? It takes some time. It’s exactly like that. The container has a whole OS in it.
Yes. It’s separating the rails+nginx into its own container and having postgres and redis in another container (or however you want to provide redis and postgres).