Reactor

Reactor is a microservice designed to test and experiment with observability of microservices. You can play with it on your own machine in local more, but it gets interesting when you deploy it on a Kubernetes cluster or an Istio mesh.

In cluster mode you can have a bit over 120 microservices (atoms in the mendeleev table). You control reactor some giving it some interesting molecules.

Motivation

I created this microservice to let me inspect what happens inside the kubernetes cluster or a mesh. I've noticed that not all the scenario's and products worked that well together, certainly when you go a bit beyond the happy path. This microservice will enable me to make reproducibles. I hope it's also helpful for other people to learn technologies like tracing, kubernetes networking, logging, istio, etc...

Reactor is for now focused on Google Cloud, as that's the cloud I live on, but once reactor is a bit more stable I'm open for other people to add other providers.

I will start using (and extending) reactor for bug reports, articles and talks. Let's see if this will get as popular as httpbin is ;).

Features

Create cascading microservice calls from one origin call
Configure tracing and logging
Inspect headers, every level in the call hierarchy

Example

To see what reactor does lets take a very simple example. Reactor works by splitting molecules:

[[H]]^2[O]

Everything between the brackets [] will be a call to the next microservice. The brackets can be prefixed with a number and an optional parameter (s or p), this tells reactor how many times the service needs to be called and how ( sequential or parallel). Multiple calls can be make by appending them using ^ (sequential) or * (parallel).

Depending the content of the bracket the call will be different. If reactor detects an atom a call to the corresponding atom service will be made. But if reactor detects another sub-molecule it calls the next orbit and apply the same logic till only atoms are left. So the example above will result in:

http://reactor-api/rr/split?molecule=[[H]]^2[O]

calling

http://orbit-1/rr/split?molecule=[H]
http://atom-o/rr/atom?atom=O
http://atom-o/rr/atom?atom=O

orbit will split the molecule [H] (ok, this looks strange, but each bracket is a layer) into it's atoms, in this case only 1 H:

http://atom-h/rr/atom?atom=H

Try the local installation, to see how it looks in the trace (this will make it more clear).

Installation

Pre-Requirement

Create a directory tmp and work in this repo. Don't worry, they are in the .gitignore so you do not accidentally check them in.

Create a Google Cloud project, enable:

Tracing
Logging
Profile
Debugging
Cloud Build
Container Registry

Replace project-name with your project name in the rest of the instructions

Create a service account for your project (eg project-name) with the appropriate roles (see above) . That will be used local as well on cluster. Download the JSON private key and copy it to work/project-name/service-account.json. Read the documentation to see how to create the service account: [https://cloud.google.com/iam/docs/creating-managing-service-account-keys].

Local

Build the reactor from source

go install ./cmd/reactor/

Set the environmental variables. Replace project-name project with your own

export PORT=3330
export REACTOR_NAME=reactor-api
export REACTOR_VERSION=1
export REACTOR_DEBUG=1
export REACTOR_PROFILE=0
export REACTOR_MODE=local
export GOOGLE_APPLICATION_CREDENTIALS=/Users/me/path/work/project-name/service-account.json
export GOOGLE_PROJECT_ID=project-name

Fire up the reactor

reactor

And test it by calling:

http://localhost:3330/rr/split?molecule=[[H]]^2[O]

Go to the Cloud Console, select Trace.

Kubernetes

Not yet fully tested/supported

gcloud builds submit --config=cb.yaml .

go install ./cmd/rrprep/

kubectl label namespace default istio-injection=disabled --overwrite

Istio

Not yet fully tested/supported

kubectl label namespace default istio-injection=enabled --overwrite

Specification

Reactor Prepare (rrprep)

Prepares the Kubernetes files, from the templates

Environment Variables

NAME	Description	Default
PORT	Port
APP_NAME	Application name
APP_VERSION	Application version
REACTOR_MODE	Reactor mode (local, k8s)	local
REACTOR_DEBUG	Debug Mode (0, 1)	1
REACTOR_PROFILE	Stackdriver Profiler (0, 1)	0
REACTOR_TRACE_PROPAGATION	OpenCensus propagator (stackdriver, b3)	stackdriver
REACTOR_TRACE_INTERNAL	Reactor adds internal spans (0, 1)	1

Molecule spec

S    [H,x=1,y=2,z=3]*2[O,x=1,y=2,z=3],x=1,y=2,z=3
A     H,x=1,y=2,z=3
A                      O,x=1,y=2,z=3

S    2[5[Ur,log:1,xyz:4]^5[C,log:1,xyz:4]],x:1,y:2
O1     5[Ur,log:1,xyz:4]^5[C,log:1,xyz:4]
A        Ur,log:1,xyz:4
A                          C,log:1,xyz:4

Test Scenarios

https://github.com/wg/wrk

wrk -t12 -c400 -d30s "http://<yourip>/rr/split?molecule=6[C]^12[H]^6[O]"

glucose 6[C]^12[H]^6[O]

https://www.sigmaaldrich.com/catalog/product/aldrich/375756?lang=en&region=BE Baicalin hydrate

C21H18O11 · xH2O

env GOOS=linux GOARCH=amd64 go build cmd/treactor/main.go

alexvanboxel / treactor-go