Overview

Pre-generate HLF crypto materials to speedup local development:

no enrollment pod
no genesis pod
orderer / peer pods start instantly

Benchmarks (local env)

	Total time to readiness	skaffold run	Deployment stabilized	Chaincode instantiated	skaffold delete
Without pre-gen	143s	100s	39s	43s	25s
With pre-gen	80s	60s	5s	20s	13s

Note: on the CI, there's a big speedup in deployment stabilization time (from ~6 min to ~2 min)

How to use

Deploy

Without pre-gen

$ skaffold run

With pre-gen

$ ./examples/secrets-create.sh
$ skaffold run

Delete

$ skaffold delete

Secrets will be deleted regardless of pre-gen status. To re-deploy with pre-gen, you need to run examples/secrets-create.sh again.

To delete secrets only, run ./examples/secrets-create.sh delete

Limitations and Future work

Skaffold deployment order is not configurable: by default, helm charts are deployed before kubectl manifest.

So if the user doesn't run examples/secrets-create.sh before skaffold run, skaffold will still create the secrets, but after deploying the helm charts. This greatly limits the speedup benefits.

There is a skaffold issue opened on the topic. Once it's addressed, we can get rid of examples/secrets-create.sh altogether.

Kelvin-M commented 4 years ago

I found ./examples/secrets-create.sh delete pretty strange, maybe it would be better to have something like :

./examples/hlf-secrets.sh create
./examples/hlf-secrets.sh delete

Kelvin-M commented 4 years ago

Moreover, we should warn the user that's only for dev setup and should not be used in production :D !

AurelienGasser commented 4 years ago

@Kelvin-M What about

./examples/sample-secrets.sh create
./examples/sample-secrets.sh delete

Would that address both your concerns?