install.sh

#!/bin/bash -eux
# The Forklift pallet github.com/PlanktoScope/pallet-standard provides the standard configuration of
# Forklift package deployments of Docker containerized applications, OS config files, and systemd
# system services for the PlanktoScope software distribution. This script integrates that pallet
# into the PlanktoScope OS's filesystem by installing Forklift and providing some systemd units
# which set up bind mounts and overlay filesystems to bootstrap the configs managed by Forklift.

config_files_root=$(dirname "$(realpath "$BASH_SOURCE")")

# Install Forklift
"$config_files_root/download-forklift.sh" "/usr/bin"

# Add the necessary systemd units (but we don't enable them until booted setup):
sudo cp "$config_files_root"/usr/lib/systemd/system/* /usr/lib/systemd/system/
sudo cp "$config_files_root"/usr/lib/systemd/system-preset/* /usr/lib/systemd/system-preset/

# Make the stage store at /var/lib/forklift/stages available for easy access in the root user's
# default Forklift workspace, both in the current boot and subsequent boots:
mkdir -p "$HOME/.local/share/forklift/stages"
sudo mkdir -p /var/lib/forklift/stages
# TODO: maybe we should instead make a new "forklift" group which owns everything in
# /var/lib/forklift?
sudo chown "$USER" /var/lib/forklift/stages
sudo systemctl enable "bind-.local-share-forklift-stages@home-$USER.service"
if ! sudo systemctl start "bind-.local-share-forklift-stages@home-$USER.service" 2>/dev/null; then
  echo "Warning: the system's Forklift stage store is not mounted to $USER's Forklift stage store."
  echo "As long as you don't touch the Forklift stage store before the next boot, this is fine."
fi

# Clone & stage a local pallet
pallet_path="$(cat "$config_files_root/forklift-pallet")"
pallet_version="$(cat "$config_files_root/forklift-pallet-version")"
forklift --stage-store /var/lib/forklift/stages plt switch --cache-img=false "$pallet_path@$pallet_version"
forklift --stage-store /var/lib/forklift/stages stage add-bundle-name factory-reset next

# Set up Forklift upgrade checks
pallet_upgrade_version_query="$(cat "$config_files_root/forklift-pallet-upgrade-version-query")"
forklift pallet set-upgrade-query "@$pallet_upgrade_version_query"

# Prepare to apply the local pallet

# Note: the pi user will only be able to run `forklift stage plan` and `forklift stage cache-img`
# without root permissions after a reboot, so we may need `sudo -E` here; I had tried running
# `newgrp docker` in the script to avoid the need for `sudo -E here`, but it doesn't work in the
# script here (even though it works after the script finishes, before rebooting):
"$config_files_root/ensure-docker.sh"
FORKLIFT="forklift --stage-store /var/lib/forklift/stages"
if ! docker info 2>&1 >/dev/null; then
  FORKLIFT="sudo -E forklift --stage-store /var/lib/forklift/stages"
fi

# Make a temporary file which may be required by some Docker Compose apps in the pallet, just so
# that those Compose apps can be successfully created (this is a rather dirty hack/workaround):
echo "setup" | sudo tee /run/machine-name

# Applying the staged pallet (i.e. making Docker instantiate all the containers) significantly
# decreases first-boot time, by up to 30 sec for github.com/PlanktoScope/pallet-standard.
if ! $FORKLIFT stage apply; then
  echo "The staged pallet couldn't be applied; we'll try again now..."
  # Reset the "apply-failed" status of the staged pallet to apply:
  $FORKLIFT stage set-next --cache-img=false next
  if ! $FORKLIFT stage apply; then
    echo "Warning: the next staged pallet could not be successfully applied. We'll try again on the next boot, since the pallet might require some files which will only be created during the next boot."
    # Reset the "apply-failed" status of the staged pallet to apply:
    $FORKLIFT stage set-next --cache-img=false next
    echo "Checking the plan for applying the staged pallet..."
    $FORKLIFT stage plan
  fi
fi

# Use forklift on future boot sequences
sudo systemctl preset forklift-apply.service
# Set up read-write filesystem overlays with forklift-managed layers for /etc and /usr
# (see https://docs.kernel.org/filesystems/overlayfs.html):
sudo systemctl preset \
  overlay-sysroot.service \
  bindro-run-forklift-stages-current.service \
  overlay-usr.service \
  overlay-etc.service \
  start-overlaid-units.service