Skip to content

Latest commit

 

History

History
90 lines (67 loc) · 3.66 KB

rundev.md

File metadata and controls

90 lines (67 loc) · 3.66 KB
Raw

Running Dataverse development snapshots

Images on Docker Hub are meant for production usage or quick demos. When developing Dataverse, testing a new feature not yet shipped in a release or running integration tests you have a need to deploy all moving parts into a (more or less) ephemeral environment.

Be warned: deployment times of Dataverse are a nightmare as of writing (Sept. 2019). For any change to the codebase, you most likely will have to wait about five to ten minutes to redeploy from compilation to reload of webpage.

This is due to a really big WAR file (which needs to be loaded into the cluster) and being stuck on old technology in combination with a tremendous monolith.

YOU HAVE BEEN WARNED.

Prepare toolchain

For efficient workflows, tools make life easier. Please install:

  1. skaffold
  2. kustomize

When you opt for using a local cluster, add:

  1. Docker
  2. kind (be nice to yourself, download a release, don't compile from source)

Note: you do have kubectl already installed, don't you?

Workflow

While you can build and deploy everything manually, it will be easier to let Skaffold take care of everything.

Running skaffold run or skaffold dev will build, tag and deploy for you. For a deeper insight, read docs at https://skaffold.dev/docs.

As you will need to access services, be sure to add --port-forward. See also port forward docs.

Before running Skaffold for the first time be sure to have a cluster at hands. You can check via kubectl. Currently only using a local cluster with kind is supported by this project, see next section. PRs welcome.

A word on waiting

Be aware that initial builds and deployments take lots of time due to cold caches. Recurring builds and deployments will be much faster, although you will still suffer from Glassfish WAR deployment times.

Typically, when there is no change to pom.xml and caches are warmed up,

  • building will take ~1 minute,
  • loading images into cluster ~45 secs,
  • deploying to K8s ~30 secs
  • and Glassfish startup + WAR deployment ~3 minutes.

How about some coffee?

Local cluster

The easiest way to work with a local cluster is using kind, which is an abbrev for "Kubernetes IN Docker".

Skaffold supports this out of the box now. Using kind, your context will be set to sth. like "@kind", which triggers loading images into a local kind cluster instead of pushing to remote registry.

After installing Docker and kind, you simply need to run:

kind create cluster

or - if your prefer a specific K8s version, e.g. 1.14.6:

kind create cluster --image kindest/node:v1.14.6

Please note that kind might have some troubles with changing networks and switching to a new DNS resolver. You might need to rebuild the cluster, which is no big deal.

If you want to use k3s, minikube, microk8s or similar, please consult the Skaffold docs, search via Google, etc.

Remote cluster

Currently only using a local cluster with kind is supported by this project. PRs welcome. Hint: we'll need Kaniko for that, as Docker Hub must not be cluttered.

Future ideas