New Book

This introduces a new version of the book.  The initial portion is built
around a tutorial building the cronjob controller.

It now uses [mdbook](https://crates.io/crates/mdbook), which is like
gitbook but maintained and written in rust.

We've got a custom "plugin" (executable) that slurps Go files into
book pages for parts of the tutorial (look for `{{#literatego
./path/to/thing}}`).

.gitignore

@@ -1,7 +1,7 @@
 .idea/
 
-# Not check in node_modules to a specific arch and os.
-docs/book/node_modules/
+# don't check in the build output of the book
+docs/book/book/
 
 # Editor temp files
 *~

docs/book/book.toml

@@ -0,0 +1,12 @@
+[book]
+authors = ["The Kubebuilder Maintainers"]
+multilingual = false
+src = "src"
+title = "The Kubebuilder Book"
+
+[output.html]
+google-analytics = "UA-119864590-1"
+curly-quotes = true
+
+[preprocessor.literatego]
+command = "./litgo.sh"

docs/book/install-and-build.sh

@@ -0,0 +1,11 @@
+#!/bin/bash
+
+os=$(go env GOOS)
+arch=$(go env GOARCH)
+
+# grab mdbook
+# mdbook's deploy got borked by a CI move, so grab our build till that gets
+# fixed (https://github.com/rust-lang-nursery/mdBook/issues/904).
+curl -sL -o /tmp/mdbook https://storage.googleapis.com/kubebuilder-build-tools/mdbook-0.2.3-${os}-${arch}
+chmod +x /tmp/mdbook
+/tmp/mdbook build

docs/book/litgo.sh

@@ -0,0 +1,11 @@
+#!/bin/bash
+
+set -ex
+
+(
+    pushd ./utils
+    go build -o ../../../bin/literate-go ./literate.go
+    popd
+) &>/dev/null
+
+../../bin/literate-go "$@"

docs/book/src/SUMMARY.md

@@ -0,0 +1,29 @@
+# Summary
+
+[Introduction](./introduction.md)
+
+[Quick Start](./quick-start.md)
+
+---
+
+- [Tutorial: Building CronJob](./cronjob-tutorial.md)
+
+  - [What's in a basic project?](./cronjob-tutorial/basic-project.md)
+  - [Every journey needs a start, every program a main](./cronjob-tutorial/empty-main.md)
+  - [Groups and Versions and Kinds, oh my!](./cronjob-tutorial/gvks.md)
+  - [Adding a new API](./cronjob-tutorial/new-api.md)
+  - [Designing an API](./cronjob-tutorial/api-design.md)
+
+      - [A Brief Aside: What's the rest of this stuff?](./cronjob-tutorial/other-api-files.md)
+
+  - [What's in a controller?](./cronjob-tutorial/controller-overview.md)
+  - [Implementing a controller](./cronjob-tutorial/controller-implementation.md)
+
+      - [You said something about main?](./cronjob-tutorial/main-revisited.md)
+
+  - [Running and deploying the controller](./cronjob-tutorial/running.md)
+  - [Epilogue](./cronjob-tutorial/epilogue.md)
+
+---
+
+[Appendix: The TODO Landing Page](./TODO.md)

docs/book/src/TODO.md

@@ -0,0 +1,7 @@
+# TODO
+
+If you're seeing this page, it's probably because something's not done in
+the book yet.  Go [see if anyone else has found
+this](https://github.com/kubernetes-sigs/kubebuilder/issues?q=is%3Aopen+is%3Aissue+label%3Akind%2Fdocumentation)
+or [bug the
+maintainers](https://github.com/kubernetes-sigs/kubebuilder/issues/new?assignees=&labels=kind%2Fdocumentation).

docs/book/src/cronjob-tutorial.md

@@ -0,0 +1,36 @@
+# Tutorial: Building CronJob
+
+Too many tutorials start out with some really contrived setup, or some toy
+application that gets the basics across, and then stalls out on the more
+complicated suff.  Instead, this tutorial should take you through (almost)
+the full gamut of complexity with Kubebuilder, starting off simple and
+building up to something pretty full-featured.
+
+Let's pretend (and sure, this is a teensy bit contrived) that we've
+finally gotten tired of the maintenance burden of the non-Kubebuilder
+implementation of the CronJob controller in Kuberntes, and we'd like to
+rewrite it using KubeBuilder.
+
+The job (no pun intended) of the *CronJob* controller is to run one-off
+tasks on the Kubernetes cluster at regular intervals.  It does the by
+bulding on top of the *Job* controller, whose task is to run one-off tasks
+once, seeing them to completion.
+
+Instead of trying to tackle rewriting the Job controller as well, we'll
+use this as an opportunity to see how to interact with external types.
+
+## Scaffolding Out Our Project
+
+As covered in the [quick start](./quick-start.md), we'll need to scaffold
+out a new project.  Make sure you've [installed
+Kubebuilder](./quick-start.md#installation), then scaffold out a new
+project:
+
+```bash
+# we'll use a domain of tutorial.kubebuilder.io,
+# so all API groups will be <group>.tutorial.kubebuilder.io.
+kubebuilder init --domain tutorial.kubebuilder.io
+```
+
+Now that we've got've a project in place, let's take a look at what
+Kubebuilder has scaffolded for us so far...

docs/book/src/cronjob-tutorial/api-design.md

@@ -0,0 +1,46 @@
+# Designing an API
+
+In Kubernetes, we have a few rules for how we design APIs. Namely, all
+serialized fields *must* be `camelCase`, so we use JSON struct tags to
+specify this.  We can also use the `omitempty` struct tag to mark that
+a field should be omitted from serialization when empty.
+
+Fields may use most of the primitive types.  Numbers are the exception:
+for API compatibility purposes, we accept two forms of numbers: `int32`
+for integers, and `resource.Quantity` for decimals.
+
+<details><summary>Hold up, what's a Quantity?</summary>
+
+Quantities are a special notation for decimal numbers that have an
+explicitly fixed representation that makes them more portable across
+machines.  You've probably noticed them when specifying resources requests
+and limits on pods in Kubernetes.
+
+They conceptually work similar to floating point numbers: they have
+a significand, base, and exponent.  Their serialize, human readable for
+uses whole numbers and suffixes to specify values much the way we describe
+computer storage.
+
+For instance, the value `2m` means `0.002` in decimal notation.  `2Ki`
+means `2048` in decimal, while `2K` means `2000` in decimal.  If we want
+to specify fractions, we switch to a suffix that lets us use a whole
+number: `2.5` is `2500m`.
+
+There are two supported bases: 10 and 2 (called decimal and binary,
+respectively).  Decimal base is indicated with "normal" SI suffixes (e.g.
+`M` and `K`), while Binary base is specified in "mebi" notation (e.g. `Mi`
+and `Ki`).  Think [megabytes vs mebibytes](../TODO.md).
+
+</details>
+
+There's one other special type that we use: `metav1.Time`.  This functions
+identically to `time.Time`, except that it has a fixed, portable
+serialization format.
+
+With that out of the way, let's take a look at what our CronJob object
+looks like!
+
+{{#literatego ./testdata/project/api/v1/cronjob_types.go}}
+
+Now that we have an API, we'll need to write a controller to actually
+implement the functionality.

docs/book/src/cronjob-tutorial/basic-project.md

@@ -0,0 +1,54 @@
+# What's in a basic project?
+
+When scaffolding out a new project, Kubebuilder provides us with a few
+basic pieces of boilerplate.
+
+## Build Infrastructure
+
+First up, basic infrastructure for building you project:
+
+<details><summary>`go.mod`: A new Go module matching our project, with basic dependencies</summary>
+
+```go
+{{#include ./testdata/project/go.mod}}
+```
+</details>
+
+<details><summary>`Makefile`: Make targets for building and deploying your controller</summary>
+```makefile
+{{#include ./testdata/project/Makefile}}
+```
+</details>
+
+<details><summary>`PROJECT`: Kubebuilder metadata for scaffolding new components</summary>
+```yaml
+{{#include ./testdata/project/PROJECT}}
+```
+</details>
+
+## Launch Configuration
+
+We also get launch configuration under the
+[`config/`](https://sigs.k8s.io/kubebuilder.io/docs/book/cronjob-tutorial/testdata/project/config)
+directory.  Right now, it just contains
+[Kustomize](https://sigs.k8s.io/kustomize) YAML definitions required to
+launch our controller on a cluster, but once we get started writing our
+controller, it'll also hold our CustomResourceDefinitions, RBAC
+configuration, and WebhookConfigurations.
+
+[`config/default`](../TODO.md) contains a [Kustomize base](../TODO.md) for launching
+the controller in a standard configuration.
+
+Each other directory contains a different piece of configuration,
+refactored out into its own base:
+
+- [`config/manager`](../TODO.md): launch your controllers as pods in the
+  cluster
+
+- [`config/rbac`](../TODO.md): permissions required to run your
+  controllers under their own service account
+
+## The Entrypoint
+
+Last, but certainly not least, Kubebuilder scaffolds out the basic
+entrypoint of our project: `main.go`.  Let's take a look at that next...

docs/book/src/cronjob-tutorial/controller-implementation.md

@@ -0,0 +1,24 @@
+# Implementing a controller
+
+The basic logic of our CronJob controller is this:
+
+1. Load the named CronJob
+
+2. List all active jobs, and update the status
+
+3. Clean up old jobs according to the history limits
+
+4. Check if we're supsended (and don't do anything else if we are)
+
+5. Get the next scheduled run 
+
+6. Run a new job if it's on schedule, not past the deadline, and not
+   blocked by our concurrency policy
+
+7. Requeue when we either see a running job (done automatically) or it's
+   time for the next scheduled run.
+
+{{#literatego ./testdata/project/controllers/cronjob_controller.go}}
+
+That was a doozy, but now we've got a working controller.  Let's test
+against the cluster, then, if we don't have any issues, deploy it!

docs/book/src/cronjob-tutorial/controller-overview.md

@@ -0,0 +1,22 @@
+# What's in a controller?
+
+Controllers are the core of Kubernetes, and of any operator.  
+
+It's a controller's job to ensure that, for any given object, the actual
+state of the world (both the cluster state, and potentially external state
+like running containers for Kubelet or loadbalancers for a cloud provider)
+matches the desired state in the object.  Each controller focuses on one
+*root* Kind, but may interact with other Kinds.
+
+We call this process *reconciling*.
+
+In controller-runtime, the logic that implements the reconciling for
+a specific kind is called a [*Reconciler*](../TODO.md).  A reconciler
+takes the name of an object, and returns whether or not we need to try
+again (e.g. in case of errors or periodic controllers, like the
+HorizontalPodAutoscaler).
+
+{{#literatego ./testdata/emptycontroller.go}}
+
+Now that we've seen the basic structure of a reconciler, let's fill out
+the logic for `CronJob`s.

docs/book/src/cronjob-tutorial/empty-main.md

@@ -0,0 +1,5 @@
+# Every journey needs a start, every program a main
+
+{{#literatego ./testdata/emptymain.go}}
+
+With that out of the way, we can get on to scaffolding our API!

docs/book/src/cronjob-tutorial/epilogue.md

@@ -0,0 +1,7 @@
+# Epilogue
+
+Things left to do:
+
+- Write custom printer columns
+- Discuss webhooks
+- Use different watches

docs/book/src/cronjob-tutorial/gvks.md

@@ -0,0 +1,69 @@
+# Groups and Versions and Kinds, oh my!
+
+Actually, before we get started with our API, we should talk terminology
+a bit.
+
+When we talk about APIs in Kubernetes, we often use 4 terms: *groups*,
+*versions*, *kinds*, and *resources*.
+
+## Groups and Versions
+
+An *API Group* in Kubernetes is simply a collection of related
+functionality.  Each group has one or more *versions*, which, as the name
+suggests, allow us to change how an API works over time.
+
+## Kinds and Resources
+
+Each API group-version contains one or more API types, which we call
+*Kinds*.  While a Kind may change forms between versions, each form must
+be able to store all the data of the other forms, somehow (we can store
+the data in fields, or in annotations).  This means that using an older
+API version won't cause newer data to be lost or corrupted.  See the
+[Kubernetes API guidelines](../TODO.md) for more information.
+
+You'll also hear mention of *resources* on occaison.  A resource is simply
+a use of a Kind in the API.  Often, there's a one-to-one mapping between
+Kinds and resources.  For instance, the `pods` resource corresponds to the
+`Pod` Kind.  However, sometimes, the same Kind may be returned by multiple
+resources.  For instance, the `Scale` Kind is returned by all scale
+subresources, like `deployments/scale` or `replicasets/scale`.  This is
+what allows the Kubernetes HorizontalPodAutoscaler to interact with
+different resources.  With CRDs, however, each Kind will correspond to
+a single resource.
+
+Notice that resources are always lowercase, and by convention are the
+lowercase form of the Kind.
+
+## So, how does that correspond to Go?
+
+When we refer to a kind in a particular group-version, we'll call it
+a *GroupVersionKind*, or GVK for short.  Same with resources and GVR. As
+we'll see shortly, each GVK corresponds to a given root Go type in
+a package.
+
+Now that we have our terminology straight, we can *actually* create our
+API!
+
+## Err, but what's that Scheme thing?
+
+The `Scheme` we saw before is simply a way to keep track of what Go type
+corresponds to a given GVK.
+
+For instance, suppose we mark that the
+`"tutorial.kubebuilder.io/api/v1".CronJob{}` type as being in the
+`batch.tutorial.kubebuilder.io/v1` API group (implicitly saying it has the
+Kind `CronJob`).
+
+Then, we can later construct a new `&CronJob{}` given some JSON from the
+API server that says
+
+```json
+{
+    "kind": "CronJob",
+    "apiVersion": "batch.tutorial.kubebuilder.io/v1",
+    ...
+}
+```
+
+or properly look up the group-version when we go to submit a `&CronJob{}`
+in an update.