It is assumed that you understand conceptually, docker containers, kubernetes, and proxy servers
*Throughout ALL documentations, this project is referred to as "ingress controller" or "the controller"
Apache Traffic Server (ATS) is a high performance, open-source, caching proxy server that is scalable and configurable. This project integrates ATS as the ingress resource to a Kubernetes(K8s) cluster, then acts as the ingress resource's custom controller.
From high-level, the ingress controller talks to K8s' API and sets up watchers
on specific resources that are interesting to ATS. Then, the controller controls ATS by either (1) relay the information from K8s API to ATS, or (2) configure ATS directly.
As defined by kubernetes/ingress-nginx:
An Ingress Controller is a daemon, deployed as a Kubernetes Pod, that watches the apiserver's
/ingresses
endpoint for updates to the Ingress resource. Its job is to satisfy requests for Ingresses.
- Alpine 3.11
- Apache Traffic Server 8.0.6
- LuaJIT 2.0.4
- Lua 5.1.4
- Go 1.12.8
- Other Packages
- luasocket 3.0rc1
- redis-lua 2.0.4
- Docker
- Kubernetes
To install Docker, visit its official page and install the correct version for your system.
The easiest way to get Kubernetes on a Mac is through Docker. Launch Docker on your machine and go to Preferences -> Kubernetes
, Click Enable Kubernetes
then Apply
. Kubernetes will be enabled and starting.
If you are cloning this project for development, visit Setting up Go-Lang for detailed guide on how to develop projects in Go.
For other purposes, you can use git clone
or directly download repository to your computer.
Once you have cloned the project repo and started Docker and Kubernetes, in the terminal:
$ cd ingress-ats
$ docker build -t ats_alpine .
- wait for Docker finish building the image
$ docker build -t node-app-1 k8s/backend/node-app-1/
- wait for Docker finish building the image
$ docker build -t node-app-2 k8s/backend/node-app-2/
- wait for Docker finish building the image
- At this point, we have created necessary images for our example. Let's talk about what each step does:
- Step 2 builds an image to create a Docker container that will contain the Apache Traffic Server (ATS) itself, the kubernetes ingress controller, along with other software required for the controller to do its job.
- Steps 3 and 4 builds 2 images that will serve as backends to kubernetes services which we will shortly create
$ kubectl create namespace trafficserver-test
- Create a namespace for ATS pod
$ openssl req -x509 -sha256 -nodes -days 365 -newkey rsa:2048 -keyout tls.key -out tls.crt -subj "/CN=atssvc/O=atssvc"
- Create a self-signed certificate
$ kubectl create secret tls tls-secret --key tls.key --cert tls.crt -n trafficserver-test --dry-run -o yaml | kubectl apply -f -
- Create a secret in the namespace just created
$ kubectl apply -f k8s/traffic-server/
- will define a new kubernetes namespace named
trafficserver-test
and deploy a single ATS pod to said namespace. The ATS pod is also where the ingress controller lives. Additionally, this will expose your local machine's port30000
to the outside world.
- will define a new kubernetes namespace named
The following steps can be executed in any order, thus list numbers are not used.
-
$ kubectl apply -f k8s/apps/
- creates namespaces
trafficserver-test-2
andtrafficserver-test-3
if not already exist - creates kubernetes services and deployments for
appsvc1
andappsvc2
- deploy 2 of each
appsvc1
, andappsvc2
pods intrafficserver-test-2
, totally 4 pods in said namespace. - similarly, deploy 2 of each
appsvc1
, andappsvc2
pods intrafficserver-test-3
, totally 4 pods in this namespace. We now have 8 pods in total for the 2 services we have created and deployed in the 2 namespaces. - in addition to the ATS pod, we have created and deployed 9 pods.
- creates namespaces
-
$ kubectl apply -f k8s/ingresses/
- creates namespaces
trafficserver-test-2
andtrafficserver-test-3
if not already exist - defines an ingress resource in both
trafficserver-test-2
andtrafficserver-test-3
- the ingress resource in
trafficserver-test-2
defines domain nametest.media.com
with/app1
and/app2
as its paths - both ingress resources define domain name
test.edge.com
; however,test.edge.com/app1
is only defined intrafficserver-test-2
andtest.edge.com/app2
is only defined intrafficserver-test-3
- Addtionally, an ingress resources defines HTTPS access for
test.edge.com/app2
in namespacetrafficserver-test-3
- creates namespaces
When both steps above have executed at least once, ATS proxying will have started to work.
In kubernetes, ingress resources are necessary to enable proxying since it is where domain names are defined. However, given only domain names, ATS cannot proxy requests when it doesn't have backend(s) to handle requests. Thus, only when both service pods and ingress are defined can ATS start proxying. To see proxy in action, we can use curl to "fake" external requests:
$ curl -vH "HOST:test.media.com" "localhost:30000/app1"
$ curl -vH "HOST:test.media.com" "localhost:30000/app2"
$ curl -vH "HOST:test.edge.com" "localhost:30000/app1"
$ curl -vH "HOST:test.edge.com" "localhost:30000/app2"
$ curl -vH "HOST:test.edge.com" -k "https://localhost:30043/app2"
With above curl commands, outputs from number 1 and 3 should be the same; outputs from number 2 and 4 should be same. The corresponding pairs are the same because all /app1
use the same backend service image, and the same goes for /app2
. Number 5 illustrates the https version for number 4 and the result is similar.
Expected received packet from /app1
resembles:
< HTTP/1.1 200 OK
< X-Powered-By: Express
< Accept-Ranges: bytes
< Cache-Control: public, max-age=0
< Last-Modified: Tue, 06 Aug 2019 16:31:53 GMT
< ETag: W/"be-16c67c5d0a8"
< Content-Type: text/html; charset=UTF-8
< Content-Length: 190
< Date: Mon, 19 Aug 2019 18:39:14 GMT
< Age: 1
< Connection: keep-alive
< Server: ATS/7.1.6
<
<!DOCTYPE html>
<HTML>
<HEAD>
<TITLE>
Hello from app1
</TITLE>
</HEAD>
<BODY>
<H1>Hi</H1>
<P>This is very minimal "hello world" HTML document.</P>
</BODY>
</HTML>
* Connection #0 to host localhost left intact
Expected received packet from /app2
resembles:
< HTTP/1.1 200 OK
< X-Powered-By: Express
< Accept-Ranges: bytes
< Cache-Control: public, max-age=0
< Last-Modified: Fri, 14 Jun 2019 18:18:51 GMT
< ETag: W/"bc-16b5736b2f8"
< Content-Type: text/html; charset=UTF-8
< Content-Length: 188
< Date: Mon, 19 Aug 2019 18:39:10 GMT
< Age: 0
< Connection: keep-alive
< Server: ATS/7.1.6
<
<!DOCTYPE html>
<HTML>
<HEAD>
<TITLE>
A Small Hello
</TITLE>
</HEAD>
<BODY>
<H1>Hi</H1>
<P>This is very minimal "hello world" HTML document.</P>
</BODY>
</HTML>
* Connection #0 to host localhost left intact
The curl commands demonstrate that, with the help of the ingress controller, ATS can not only resolve domain names while routing requests to various namespaces based on path, but also, is capable of handling the case where domain's paths existing across different namespaces.
Of course there are checks in place by the ingress controller so that any path corresponding to a domain name can only be defined in one namespace, and domain names are resolved across all namespace.
Below is an example of configuring Apache Traffic Server reloadable configurations using kubernetes configmap resource:
$ kubectl apply -f k8s/configmaps/
- create a ConfigMap resource in
trafficserver-test
if not already exist - configure 3 reloadable ATS configurations:
proxy.config.output.logfile.rolling_enabled: "1"
proxy.config.output.logfile.rolling_interval_sec: "3000"
proxy.config.restart.active_client_threshold: "0"
- feel free to add other reloadable configurations, and/or change the above 3 to other valid values. Checks are in place so that mistakes are tolerated.
- create a ConfigMap resource in
You can attach ATS lua script to an ingress object and ATS will execute it for requests matching the routing rules defined in the ingress object. See an example in k8s/ingresses/ats-ingress-2.yaml
You can provide an environment variable called INGRESS_CLASS
in the deployment to specify the ingress class. Only ingress object with annotation kubernetes.io/ingress.class
with value equal to the environment variable value will be used by ATS for routing
- Get Go-lang 1.12 from official site
- Add
go
command to your PATH:export PATH=$PATH:/usr/local/go/bin
- Define GOPATH:
export GOPATH=$(go env GOPATH)
- Add Go workspace to your PATH:
export PATH=$PATH:$(go env GOPATH)/bin
- Define Go import Paths
- Go's import path is different from other languages in that all import paths are absolute paths. Due to this reason, it is important to set up your project paths correctly
- define the base path:
mkdir -p $GOPATH/src/github.com/<your user name>/
- Clone the project:
cd $GOPATH/src/github.com/<your user name>/
git clone <project>
- As of Go 1.12 in order to have
go.mod
within Go paths, you must export:export GO111MODULE=on
to be able to compile locally.
*Above steps are a very short summary of Getting Started and How to Write Go Code from official Go-lang documentation. For more detailed info and/or assistance, it is recommended to checkout these 2 links first.
To compile, while in ingress-ats/
directory: go build -o ingress_ats main/main.go
The repository comes with basic support for both vscode and vim
.
If you're using vscode
:
.vscode/settings.json
contains some basic settings for whitespaces and tabs.vscode/extensions.json
contains a few recommended extensions for this project. It is highly recommended to install the Go extension since it contains the code lint this project used during development.
If you're using vim
, a vimrc
file with basic whitespace and tab configurations is also provided