rstats-ubuntu

R template for OpenFaaS with classic watchdog based on rocker/r-ubuntu

Sends and receives JSON using classic watchdog.

This template uses the rocker/r-ubuntu:18.04 image and the repos option to Ubuntu bionic via RSPM.

Making a new function

Use the faas-cli and pull R templates

faas-cli template pull https://github.com/analythium/openfaas-rstats-templates

Use the rstats-ubuntu function template and create a new function called r-ubuntu-hello; prefix the dockeruser to the docker image tag (i.e. dockeruser/r-ubuntu-hello will be the image name):

faas-cli new --lang rstats-ubuntu r-ubuntu-hello --prefix=dockeruser

Customizing your function

Now we have a r-ubuntu-hello.yml file and function folder ./r-ubuntu-hello. Files in the function folder will get copied to the /home/app directory of the image. Read more about the YAML configuration. Customize the ./r-ubuntu-hello/handler.R file as needed:

• load required packages using library(),
• put your data in the folder and load it relative to the function folder (e.g. data.RData) or use the full path (e.g. /home/app/data.csv),
• define the output given the input of the handle function, the output must be JSON parsable,
• add packages/remotes/system requirements and optionally metadata to the DESCRIPTION file.

Build, push, and deploy

The up command includes build (build an image into the local Docker library) push (push that image to a remote container registry), and deploy (deploy your function into a cluster):

faas-cli up -f r-ubuntu-hello.yml

Now you should see something like this:

[0] > Building r-ubuntu-hello.
...
[0] < Building r-ubuntu-hello done in 10.97s.
[0] Worker done.

Total build time: 0.97s

[0] > Pushing r-ubuntu-hello [dockeruser/r-ubuntu-hello:latest].
...
[0] < Pushing r-ubuntu-hello [dockeruser/r-ubuntu-hello:latest] done.
[0] Worker done.

Deploying: r-ubuntu-hello.
WARNING! Communication is not secure, please consider using HTTPS. Letsencrypt.org offers free SSL/TLS certificates.

Deployed. 202 Accepted.
URL: http://IP_ADDRESS:8080/function/r-ubuntu-hello.openfaas-fn

Testing

Test the local Docker image forwarding to port 4000

docker run -p 4000:8080 dockeruser/r-ubuntu-hello

Curl should return ["Hello Friend!"]:

curl http://localhost:4000/ -d '["Friend"]'
# ["Hello Friend!"]

The log looks like this:

2020/12/12 04:25:18 Version: 0.20.1     SHA: 7b6cc60bd9865852cd11c98d4420752815052918
2020/12/12 04:25:18 Timeouts: read: 5s, write: 5s hard: 0s.
2020/12/12 04:25:18 Listening on port: 8080
2020/12/12 04:25:18 Metrics listening on port: 8081
2020/12/12 04:25:18 Writing lock-file to: /tmp/.lock
2020/12/12 04:27:56 Forking fprocess.
2020/12/12 04:27:56 Wrote 18 Bytes - Duration: 0.335202s
...

Use the OpenFaaS UI or curl (should give ["Hello Friend!"]). Replace localhost with IP address if testing on remote location:

curl http://localhost:8080/function/<function-name> -d '["Friend"]'
# ["Hello Friend!"]

Example: PCA

Create a new function:

faas-cli new --lang rstats-ubuntu r-ubuntu-pca --prefix=dockeruser

Change handler.R:

suppressMessages(library(vegan))
handle <- function(req) {
    scores(rda(req), 1:2)$sites
}

Edit DESCRIPTION:

Package: OpenFaaStR
Version: 0.0.1
Imports:
  vegan
Remotes:
  vegandevs/vegan
SystemRequirements:
  libgfortran3,
  libgfortran5
VersionedPackages:

Note: need to build vegan from source otherwise it cannot link to shared libraries. Therefore we add it to Remotes (we have to list it under Imports, but Remotes indicates that it is to be installed from GitHub and not from the RSPM CRAN repo).

Build the image: faas-cli build -f r-ubuntu-pca.yml and test with docker run -p 4000:8080 dockeruser/r-ubuntu-pca and

curl http://localhost:4000/ -H \
  "Content-Type: application/json" -d \
  '[[-1,3,16],[10,-10,9],[-5,10,-14],[14,3,-12]]'
# [[2.9545,3.042],[2.7754,-2.5023],[-3.4647,2.7223],[-2.2652,-3.262]]