Open Match User Experience TODOs

[yfei1]

Open this issue to track the Eng and Doc efforts to improve Open Match's User Experience. I'll break down the overall user experience into several pieces with different prioritization levels.

: P0, need to be done in the next three weeks : P1, need to be done in six weeks : P2, need to be done before 1.0

Basic: How to do match making with Open Match?

This part covers why, what, and how to implement Open Match customized components.

• Doc:
• [x] *Have quick starts that explain why do we need the Director, MMF, and Evaluator.
• [x] *Have quick starts that show users how to implement these components.
• [x] *Have a brief README under the examples/ directory to explain which subdirectory users should look into.
• Eng:
• [x] *Refactor current examples to use public libraries.
• [x] *Have a second-match demo installed in a different namespace

Basic: How to build and deploy Open Match?

It would be great if Open Match can provide some tools/scripts to build and deploy Open Match to a k8s cluster without having users writing the Dockerfiles, yamls for k8s resources.

• Doc:
• [ ] **Provide docs to show how to use the user Makefile or shell script.
• [x] *Provide docs to show how to deploy mmf and evaluator with minimalistic settings.
• [x] *Explain what were installed when users ran kubectl apply -f install.yaml
• Eng:
• [x] *Have the quick start only install open-match-core components - Jaeger, grafana, prometheus... are too much to cover for a quick start, perhaps we should have telemetry supports installed only in the advanced guide?
• [ ] **Have a unified Dockerfile to build and push images for the customized components - we should also have a separate Makefile or shell script for the Open Match users - This would likely come after v0.8, current plan is to have all the example codes checked in and discuss where they should be placed in the end.
• [x] *Have minimalistic pod definitions in the examples/resources directory to deploy the MMF and Evaluator using cat mmf.yaml | kubectl apply -n open-match -f -. Note that this would only deploy a pod, simply to show how to have the customized components up and running.

Basic: How to check if Open Match works as expected with the customized components?

Have a fake service that will generate tickets and fetch matches under both k8s and local environment.

• Doc:
• [ ] **Provide docs to show how to run the fake service with one line - e.g. ideally just do go run fake_service.go.
• [x] **Provide docs to show the expected logs from Open Match and customized components after running the fake service.
• Eng:
• [x] *Have demo's director calling FetchMatches with a pool of no filters, and wrap demo's frontend and director as a service.
• [ ] **Refactor minimatch into a service/Have a new CLI service without dependencies on local config files.

***Advanced: How to customize Open Match and deploy scalable MMF/Evaluators using helm?

Ideally P1, depends on the release date of helm3

• [x] Push Open Match's helm chart to helm hub.
• [x] Have a doc for how to customize Open Match with helm using its configurable parameters.
• [x] Have a template for scalable deployments of MMF & Evaluator
• [ ] Guide on how to install Linkerd/do gRPC client side load balancing

Advanced: How to customize and use Open Match's telemetry supports?

• [x] Docs: **How to use Prometheus, Grafana, Jaeger, Zipkin, zPages, Stackdriver with Open Match
• [ ] Docs: ***If you have an existing telemetry system, how to integrate it with Open Match?
• [x] Eng: *Add more internal metrics to Open Match. Open Match now has built-in metrics for gRPC calls, but lack of metrics for internal Redis calls.
• [x] Eng: **Integrate Jaeger with Open Match.
• [ ] Eng: ***Template configurations of the telemetry supports to enable integrations with existing telemetry system.

***Advanced: How to do real time scaling/rolling update with Open Match?

[sawagh]

Please revisit this to identify a similar todo list for v0.9

[sawagh]

@yfei1 - Please review the remaining tasks here. I would prefer having separate issues for pending tasks - than longer lived issues - easier to pin them to milestones.

[sawagh]

Closing this as we have reasonable UX tasks created. Please create separate tasks for new issues if any.