Error while deploying R using plumber in Google App Engine Flex with Docker

Question

What am doing?

I am trying to deploy a R model on Google App Engine Flex with docker container. My final objective is to serve model as API. I am getting errors when deploying app using plumber and docker container.

R code with plumber runs fine on my local computer using RStudio. But now I am using AI platform jupyter notebooks with R. I tested the docker locally using Docker Run image-name command and I get below message once Docker run.

Starting server to listen on port 8080 

When I run the R + plumber code in my local Rstudio , I get below messages

Starting server to listen on port 8080
Running the swagger UI at http://127.0.0.1:8080/__swagger__/

After this I run gcloud app deploy ( this agains build docker image etc) , build runs for more than 15 mins and fails with error message , as shown in the end.

Details of code etc:

app.yaml

service: iris-custom
runtime: custom
env: flex

manual_scaling:
  instances: 1

resources:
  cpu: 1
  memory_gb: 0.5
  disk_size_gb: 20

# added below to increase app_start_timeout_sec  
readiness_check:
  path: "/readiness_check"
  check_interval_sec: 5
  timeout_sec: 4
  failure_threshold: 2
  success_threshold: 2
  app_start_timeout_sec: 900

Dockerfile

FROM gcr.io/gcer-public/plumber-appengine

# install the linux libraries needed for plumber
RUN export DEBIAN_FRONTEND=noninteractive; apt-get -y update \
&& apt-get install -y

# install plumber commented as plumber is preinstalled
#RUN R -e "install.packages(c('plumber'), repos='http://cran.rstudio.com/')"

# copy everything from the current directory into the container
WORKDIR /payload/
COPY [".", "./"]

# open port 8080 to traffic
EXPOSE 8080

# when the container starts, start the main.R script
ENTRYPOINT ["Rscript", "main.R"]

main.R

library(plumber)
r <- plumb("rest_controller.R")
r$run(port=8080, host="0.0.0.0")

rest_controller.R

#* @get /predict_petal_length
get_predict_length <- function(){

  dataset <- iris

  # create the model
  model <- lm(Petal.Length ~ Petal.Width, data = dataset)
  petal_width = "0.4"

  # convert the input to a number
  petal_width <- as.numeric(petal_width)

  #create the prediction data frame
  prediction_data <- data.frame(Petal.Width=petal_width)

  # create the prediction
  predict(model,prediction_data)
}

Error message:

ERROR: (gcloud.app.deploy) Error Response: [4] Your deployment has failed to become healthy in the allotted time and therefore was rolled back. If you believe this was an error, try adjusting the 'app_start_timeout_sec' setting in the 'readiness_check' section.

I tried a little modified code ,deployment succeeds but app engine still does not work. issue with code link

Answer 1

From the Google Cloud Doku it seems like in order for your Apllication to pass it needs to return the http status code 200 (see https://cloud.google.com/appengine/docs/flexible/custom-runtimes/configuring-your-app-with-app-yaml#updated_health_checks ).

But your Application returns the http status code 404 on the path you have defined for redincess check, since it doesn't exist.

readiness_check:
 path: "/readiness_check"

So I would either suggest to add this path as an option to your rest_controller.R file like

#* @get /readiness_check
readiness_check<- function(){
    return ("app ready")
}

or modify your app.yml so that it checks the get_predict_length enpoint instead

readiness_check:
  path: "/get_predict_length"