docker-digitalocean-furrr.R

library(tidyverse)
library(analogsea)
library(furrr)

# R on local machine → DigitalOcean Linux image → Docker on that → R

# Install analogsea, log into DigitalOcean, generate an API key with write
# permissions, and add this line to ~/.Rprofile:
#
# Sys.setenv(DO_PAT = "KEY_GOES_HERE")

# Create a droplet with Docker pre-installed
d <- docklet_create(region = "sfo2")

# Pull the latest tidyverse image onto the droplet
droplet(d$id) %>% docklet_pull("rocker/tidyverse")

# Get the public IP address for the droplet
ip <- d$networks$v4[[1]]$ip_address

# Path to your private key (make sure your key is in your DigitalOcean account too)
ssh_private_key_file <- "/Users/andrew/.ssh/id_rsa"

# R script that needs to run a future command
# It will run the tidyverse image; --net=host allows it to communicate back to this computer
rscript <- c("sudo", "docker", "run", "--net=host", 
             "rocker/tidyverse", "Rscript")

# Connect and create a cluster
cl <- makeClusterPSOCK(
  
  # Public IP for droplet(s); this can be a vector of IP addresses
  ip,
  
  # User name; DO droplets use root by default
  user = "root",
  
  # Use private SSH key registered with DO
  # (I actually have no idea if StrictHostKeyChecking and IdentitiesOnly are
  # necessary—they're from a different tutorial about doing this with AWS.)
  # ¯\_(ツ)_/¯
  rshopts = c(
    "-o", "StrictHostKeyChecking=no",
    "-o", "IdentitiesOnly=yes",
    "-i", ssh_private_key_file
  ),
  
  # Script to run
  rscript = rscript,
  
  # Set up .libPaths() for the root user and install future/purrr/furrr packages
  rscript_args = c(
    "-e", shQuote("local({p <- Sys.getenv('R_LIBS_USER'); dir.create(p, recursive = TRUE, showWarnings = FALSE); .libPaths(p)})"),
    "-e", shQuote("if (!requireNamespace('furrr', quietly = TRUE)) install.packages('furrr')")
  ),
  
  dryrun = FALSE
)

# Use the cluster!
plan(cluster, workers = cl)
future_map(1:5, ~.x)