search

concourse / concourse-bosh-deployment

A toolchain for deploying Concourse with BOSH.
Apache License 2.0
86 stars 155 forks source link

Single vm #133

Closed jaimegag closed 5 years ago

jaimegag commented 5 years ago

Changes in this PR include an ops file (single-vm.yml) that collapses the 3 instance-groups of the cluster deployment into 1. The purpose is to enable users to deploy a single-vm Concourse with an existing Bosh Director, primarily targeting BBL users that don't want a large Concourse deployment

To facilitate reuse of most existing ops files, the remaining instance-group will be called web.

On the other hand, since all jobs now run in that same instance-group, this single-vm must be sized accordingly and must incorporate the db_persistent_disk. Additionally, for many pipelines that manage large artifacts (i.e: pcf-pipelines) you need a larger ephemeral disk. For this purpose I have included a second ops-file (web-ephemeral-disk.yml).

vito commented 5 years ago

Hmm sorry but I'm not sure about this either. It duplicates parts of the original manifest, and the way it supports ops files is kind of clunky (only supports ops files that update web and names the whole instance group web when really it's everything).

I don't think ops files are powerful enough for this kind of thing. It might be better to just have a script or executable of some sort that can take a multi-instance group manifest and output a manifest that has them all in one instance group.

Sorry to be such a stickler.

jaimegag commented 5 years ago

@vito

I think that building a tool that uses any possible multi-instance-group manifest as input and outputs a single-instance-group manifest would be great but it could end up being quite complex if you want to handle all different combinations resulting from existing different ops-files (and there are many). I certainly don't have the cycles to create that.

What I built is far from perfect; yes, it does duplicate some things, and limits reusability of ops-files to the ones changing the web instance group. But it still can help on quite a few scenarios, it does not add much clutter to the repo (2 independent ops-files) ... and I personally think it is better than not having anything else like this!.

No worries if you rather keep this out of the repo. I'll keep it in a branch in my forked version.

taylorsilva commented 5 years ago

I'm going to close this issue because it has gone stale.