Docker: MySQL refuses host IP unless I restart it manually - mysql

Question

Here is my Dockerfile:
FROM debian:jessie-backports
RUN apt-get update --yes && apt-get upgrade --yes
RUN DEBIAN_FRONTEND=noninteractive apt-get install -y --no-install-recommends \
git \
curl \
build-essential\
libssl-dev\
ca-certificates\
mysql-server\
redis-server\
elasticsearch
USER root
ENV HOME /root
# MYSQL SETUP
RUN sed -i -e"s/^bind-address\s*=\s*127.0.0.1/#bind-address = 0.0.0.0/" /etc/mysql/my.cnf
RUN sed -i -e"s/^#max_connections\s*=\s*100/max_connections = 200/" /etc/mysql/my.cnf
RUN echo "\n[mysqld]\nskip-grant-tables\n" >> /etc/mysql/my.cnf
VOLUME ["/var/lib/mysql", "/var/log/mysql"]
EXPOSE 3306
# REDIS SETUP
RUN echo "daemonize yes\nbind 0.0.0.0" >> /etc/redis/redis-serve.conf
RUN sed 's/^daemonize no/daemonize yes/' -i /etc/redis/redis.conf \
&& sed 's/^bind 127.0.0.1/bind 0.0.0.0/' -i /etc/redis/redis.conf \
&& sed 's/^# unixsocket /unixsocket /' -i /etc/redis/redis.conf \
&& sed 's/^# unixsocketperm 755/unixsocketperm 777/' -i /etc/redis/redis.conf \
&& sed '/^logfile/d' -i /etc/redis/redis.conf
VOLUME ["/var/lib/redis", "/var/log/redis"]
EXPOSE 6379 6380
# ELASTICSEARCH SETUP
RUN sed 's/^#START_DAEMON=true/START_DAEMON=true/' -i /etc/default/elasticsearch
VOLUME ["/opt/elasticsearch/data", "/opt/elasticsearch/logs"]
EXPOSE 9200 9300
ADD docker-entrypoint.sh /usr/local/bin/docker-entrypoint.sh
RUN chmod 777 /usr/local/bin/docker-entrypoint.sh
ENTRYPOINT ["/usr/local/bin/docker-entrypoint.sh"]
The entrypoint:
#!/bin/bash
/etc/init.d/mysql start
/etc/init.d/redis-server start
/etc/init.d/elasticsearch start
/bin/bash
And the command I'm running to start the container:
docker run -i -t -p 3306:3306 -p 6379:6379 -p 9200:9200 -p 9300:9300 assethost
I want to connect to the MySQL server running in the container from the host, but this is what happens when I try to connect:
ERROR 1130 (HY000): Host '192.168.99.1' is not allowed to connect to this MySQL server
Both Redis and Elasticsearch expose themselves just fine. MySQL will only begin accepting external hosts if I restart it manually from the Bash shell that comes up from the entrypoint when I run the container. I have to run /etc/init.d/mysql restart. Only then, can I successfully connect from the host using the MySQL client.
Note that I am using docker-machine, so I have to give the MySQL client the host IP in order to connect.
How can I be able to connect from the host to the MySQL instance running in the container without having to manually restart MySQL?
Docker version 1.12.1, build 23cf638

Answer 1

I figured out a way to get the skip-grant-tables option to be applied on start.
Workaround
In docker-entrypoint.sh, I changed
/etc/init.d/mysql start to
HOME=/etc/mysql /usr/bin/mysqld_safe > /dev/null 2>&1 &
Other changes
Since there was already a [mysqld] section in my.cnf, I changed the following line in Dockerfile
RUN echo "\n[mysqld]\nskip-grant-tables\n" >> /etc/mysql/my.cnf
to
RUN sed -i '/\[mysqld\]/ a skip-grant-tables\nskip-name-resolve' /etc/mysql/my.cnf
I added skip-name-resolve so mysql doesn't attempt to resolve names for local ip addresses.
Caveats
The workaround no longer works when /etc/init.d/mysql restart is executed, but /etc/init.d/mysql stop; /etc/init.d/mysql start preserved the 'skip-grant-tables' option.

Answer 2

I'm guessing MySQL has already started after your installation and running sed to change configuration won't take effect. Entry point you have has mysql start but it is probably seeing that MySQL is already running and not doing anything. Try changing that command to mysql restart, so your entry point bash script becomes:
#!/bin/bash
/etc/init.d/mysql restart
/etc/init.d/redis-server start
/etc/init.d/elasticsearch start
/bin/bash

Answer 3

You're using skip-grant-tables flag to login without password. I think due to some reasons of /etc/init.d/mysql script, the flag is ignored on the first start, making the connection from the root#192.168.99.1 rejected. I believe Docker has nothing to do with this issue.
To answer your question, you don't have to restart by hand, you can change your entrypoint to either:
#!/bin/bash
/etc/init.d/mysql start
# Wait for mysql to be fully up, then restart (e.g., 5 seconds)
sleep 5
/etc/init.d/mysql restart
/etc/init.d/redis-server start
/etc/init.d/elasticsearch start
/bin/bash
Or, don't use /etc/init.d/mysql start, instead invoke mysqld directly
#!/bin/bash
mysqld &
/etc/init.d/redis-server start
/etc/init.d/elasticsearch start
/bin/bash
In addition, although creating one more [mysqld] group in my.cnf is not an issue, you should append new configuration to the existing group:
RUN sed -i '/\[mysqld\]/a skip-grant-tables' /etc/mysql/my.cnf
Refer the doc for several ways to start/stop Mysql server