A better wait-for-postgres.sh

wait-for-postgres.sh

#!/bin/bash -e
# wait-for-postgres.sh
# Adapted from https://docs.docker.com/compose/startup-order/

# Expects the necessary PG* variables.

until psql -c '\l'; do
  echo >&2 "$(date +%Y%m%dt%H%M%S) Postgres is unavailable - sleeping"
  sleep 1
done
echo >&2 "$(date +%Y%m%dt%H%M%S) Postgres is up - executing command"

exec ${@}

here's better variant

#!/bin/sh -e

# MIT
# idea stolen from https://gist.github.com/nicerobot/1136dcfba6ce3da67ce3ded5101a4078

# USAGE
# DBNAME=postgres://... ./wait-for-postgres

# Retries a command on failure (idea stolen from http://fahdshariff.blogspot.com/2014/02/retrying-commands-in-shell-scripts.html).
# $1 - the max number of attempts
# $2 - the seconds to sleep
# $3... - the command to run
retry() {
  max_attempts="$1"; shift
  seconds="$1"; shift
  cmd="$@"
  attempt_num=1

  until $cmd
  do
    if [ $attempt_num -eq $max_attempts ]
    then
      echo "Attempt $attempt_num failed and there are no more attempts left!"
      return 1
    else
      echo "Attempt $attempt_num failed! Trying again in $seconds seconds..."
      attempt_num=`expr "$attempt_num" + 1`
      sleep "$seconds"
    fi
  done
}

retry 5 1 psql --dbname=$DBNAME -c '\l' >/dev/null

echo >&2 "$(date +%Y%m%dt%H%M%S) Postgres is up - executing command"

exec ${@}

"Better" is subjective. It's more versatile maybe. That doesn't always mean better ;-) Simplicity also matters and I didn't need a retry limit.

Also, I wouldn't provide the --dbname flag. psql supports PG* environment variables. Let the caller provide them in the way psql already expects.

I would also adhere to a different Bash style.

So if you want retries and want a truly better variant, I would write it like this:

#!/usr/bin/env bash

set -o errexit
set -o nounset
set -o pipefail

retry() {
  local max_attempts="${1}"; shift
  local retry_delay_seconds="${1}"; shift
  local cmd="${@}"
  local attempt_num=1

  until ${cmd}; do
    (( attempt_num >= max_attempts )) && {
      echo "Attempt ${attempt_num} failed and there are no more attempts left!"
      return 1
    }
    echo "Attempt ${attempt_num} failed! Trying again in ${retry_delay_seconds} seconds..."
    attempt_num=$[ attempt_num + 1 ]
    sleep ${retry_delay_seconds}
  done
}

retry 1>&2 ${MAX_ATTEMPTS:-5} ${RETRY_DELAY_SECONDS:-1} psql -c '\l'

psql "${@}"

This example (as the original one in the docker website) is using the psql command. This command is not available in the "webapp" service though (only available in the "db" service). Am I getting something wrong?

You will need to install the "postgresql-client" package for these commands to be available in the webapp.

The included pg_isready command achieves the same as above without the retry limit.

See the complete config using this technique here.

A simplified Dockerfile is shown below. You will need to include a pgsql link for hostname in docker-compose.yml.

FROM openjdk:11

RUN apt-get update && apt-get install -y procps curl net-tools telnet
RUN apt-get install -y postgresql-client

#other things

COPY build/libs/app-0.0.1.war app.war

EXPOSE 8080

CMD until pg_isready --host=pgsql; do sleep 1; done \
    && java -jar app.war

For the above script replace the CMD with :

CMD wait-for-postgres.sh && java -jar app.war