to give instructions for potential contributor

For other contributor

We need to turn on health check plugin in the image and then expose the support in helm chart as well.

Need to compare and measure the slowdown of using redis for doing routing instead of using remap.config rules.

Instead we should get the informer from factory

ATS can support proxy-wasm
There is a POC we can build on top of - https://github.com/jplevyak/trafficserver/tree/wasm

We should also check out this presentation - https://static.sched.com/hosted_files/kccnceu20/95/building%20a%20distributed%20API%20Gateway%20with%20a%20service%20mesh.pdf for inspiration to provide webassembly scripting support to the ATS ingress controller.

Go1.16 is released
https://blog.golang.org/go1.16

See here - https://kubernetes.io/blog/2020/04/02/improvements-to-the-ingress-api-in-kubernetes-1.18/
for details

To install the ATS ingress controller

https://docs.trafficserver.apache.org/en/8.1.x/admin-guide/configuration/hierachical-caching.en.html

For routing from ATS to the pod.
Right now it is only doing it randomly. We can add support for round-robin and others

We are running as root now. We should change that.

If we don't do crond/syslogd as well as the related log rotation for the go program, we should be able to run as nobody

"Actions in this workflow must be: created by GitHub, verified in the GitHub Marketplace, within a repository owned by apache or match the following: adoptopenjdk/, apache/, gradle/wrapper-validation-action."

So we are using

• leafo/[email protected]
• leafo/[email protected]
• manusa/[email protected]

Need to follow

• https://issues.apache.org/jira/browse/INFRA-21239
• https://issues.apache.org/jira/browse/INFRA-21234

And see how to get these allowed again.

Update all actions to use the following

actions/checkout@v4
actions/setup-go@v5

And stop using set-output command - see https://hynek.me/til/set-output-deprecation-github-actions/

Upgrade codeql action to v3 - https://github.blog/changelog/2024-01-12-code-scanning-deprecation-of-codeql-action-v2/

e.g. namespace of the ingress object and the ignoreNamespace

We can add them through environment variable. Similar to ingress class.

First problem is that libexecinfo is not available with alpine linux 3.17+

It is currently 0. We should set it to 30 seconds and test it out.

These features are configuration in https://docs.trafficserver.apache.org/en/8.0.x/admin-guide/files/ssl_server_name.yaml.en.html
can use annotation of the ingress object to specify these.
e.g.
https://kubernetes.github.io/ingress-nginx/user-guide/nginx-configuration/annotations/#client-certificate-authentication
https://kubernetes.github.io/ingress-nginx/user-guide/nginx-configuration/annotations/#backend-certificate-authentication
https://kubernetes.github.io/ingress-nginx/user-guide/nginx-configuration/annotations/#ssl-passthrough

Similar to https://github.com/haproxytech/helm-charts/tree/d48d40ed9a97fd23158405fe3f356af94d4049c4/kubernetes-ingress

Pod security policy
Ingress class
HPA / KEDA
PodDisruptionBudget:
ConfigMap

Right now the integration tests do not test anything being added to the helm chart, we should add integration tests that use helm install

I would like to test the ingress controller in a production setting.

i.e. with frequent changes in backend application deployment, scale up/down and see if the ingress controller can indeed handle all the changes and can fail gracefully as well.

1.27 End of life is 06/28/2024

Should incorporate this - github/roadmap#918

Let's use Jaegar to illustrate the support

If there is failure writing to redis, we should consider using a work queue to retry the event later.

We should also consider other measure to improve resiliency of the ingress controller under the pressure of high volume updates.

Since config maps for fluentd and ats reside in the same namespace, configmap handler picks up the config maps added for fluentd and tries to execute the command 'traffic_ctl set config' with these configmaps

More information here - https://blog.openresty.com/en/luajit-gc64-mode/

Similar to what nginx ingress controller can do - https://dev.to/bzon/nginx-ingress-load-balancing-and-retry-1008

We need to investigate whether we can use similar mechanism as the escalate plugin - https://github.com/apache/trafficserver/tree/master/plugins/escalate for this purpose

As a first version, we can set up ramdisk to be used as cache storage for ATS

Similar to https://kubernetes.github.io/ingress-nginx/kubectl-plugin/