Scott Johnston

Docker 0.11 is the Release Candidate for 1.0

Scott Johnston

One of the most-asked questions we’ve been hearing these last couple months is, “When is 1.0 coming?”  While not a direct answer, today we’re pleased to announce the immediate availability of Docker 0.11 – the first Release Candidate (RC) for 1.0.  From here, we want to continue to get your feedback, focus on quality.

RC designation aside, there are a lot of features to be excited about in this release.  Here are the big ones:

  • SELinux support.  Big shout-out to community member Dan Walsh, “Mr. SELinux” of Red Hat fame, for the work he contributed on this. To run the daemon with SELinux support, use docker -d --selinux-enabled

  • Host networking.  For those wanting their containers to directly access the host system’s network interfaces, Docker 0.11 features a “host networking” mode. A container in host networking mode is sandboxed as usual, except for its network stack which is shared with the host system.

    By default, when you start a container, it’ll only have a loopback interface `lo` and a virtual interface `eth1` conected to the docker bridge:

    $> docker run busybox ip a

    You can now directly use the interfaces of you host (it’s much faster as it doesn’t go through the bridge) via `–net=host`:

    $> docker run --net=host busybox ip a
  • Link hostnames. When containers are linked together, they can now discover each other by hostname. For example your application frontend could connect to its database by simply opening a connection to the hostname “db”. This discovery method is optional, you can still use the usual environment variable discovery method.You can start a redis server on you machine, giving it the name ‘redis’ with:

    $> docker run -d --name redis crosbymichael/redis

    and if you start a redis-cli container with the `redis` container linked into it, you can use:

    $> docker run -it --name redis-cli --link redis:redis crosbymichael/redis-cli -h redis

    as the `redis-cli` container will have it’s /etc/hosts populated the `redis` container ip.

  • Ping Docker daemon.  We’ve added an endpoint that’ll allow you to ping the Docker daemon to see if it’s alive for things like health checks.

  • Timestamps.  With Docker 0.11, there are now timestamps in all the logs provided by the daemon for each container. See docker logs -t

  • Access registry mirrors with multiple endpoints.  This feature in the header of Docker calls allows for chaining of multiple registry mirrors, eg, if one mirror has an outage, re-direct to the standby mirror, etc.

  • SHA-512.  Adding SHA-512 enables support for certain SSL certificates.

The above major features plus bug fixes total 266 merged pull requests in Docker 0.11 – check out the full list.  Big thanks and shout-out to all contributors, particularly Dan Walsh, Victor Marmol, and Rohit Jnagal!

As we head into the home stretch toward 1.0, we look forward to your feedback and input!

Dockerize early and often,

– The Docker Team

Learn More


24 thoughts on “Docker 0.11 is the Release Candidate for 1.0

  1. Thiago Martins

    Cool!! But, please, add native support for IPv6.

  2. Link Hostnames is cool but doesn’t quite cover the use-case of discovering hosts when they’re launched post creation of host that needs to discover.

    Was hoping a feature would eventually come out that would eliminate the need for “crosbymichael/skydock”, unfortunately link hostnames feature is not it.

  3. Katherine Crowson

    GPG signed builds please?

    • James Turnbull

      James Turnbull

      We’ll be providing these shortly but in the meantime we do issue SHA/hashes for the binaries in the meantime.

  4. Amazing work guys. I’ve spent the week moving our production systems to docker, and it’s been a blast.

  5. Great! But the Mac OS client is still 0.10. 🙁

    • James Turnbull

      James Turnbull

      David – It’s been updated now. In the past it’s usually a couple of hours behind the release but we’ve just updated our release process to ensure we’ll be shipping new Boot2Docker versions as part of the release.

  6. New links is just awesome. It’s like a magic dream after parsing environments.
    Host networking also is very promising.

  7. Please note that if you specify that you want to use the lxc container engine by default (e.g. adding -DOCKER_OPTS=”-e lxc” to /etc/default/docker), then –net=host results in you only having the loopback interface.

  8. Nice 😉

  9. docker run –net=host busybox ip a
    2014/05/09 08:14:22 unable to remount sys readonly: device or resource busy

    It not works.

  10. That hostname link feature is probably the best feature I’ve ever seen, can’t wait to play with it!

  11. Michel Peterson

    This is great! Thanks to the host network links it allowed me to create a container that I was meaning for a long time:

  12. I dig the host networking, but would be ecstatic if I could specify which device(s) I wanted to bring from the host to the container. I’ve been using DNAT and SNAT on the host to kind of get this configuration, but being able to do something like -net=”virtual0″ where virtual0 is a macvlan interface would totally sweet.

    Keep up the great work!

  13. Great work guys!
    I see an rpm package in the fedora20 updates testing repo but not in the fedora 19.
    There are any compatibility issues between 0.11 and fc19 ?


  14. When I run docker with –net=host, run command “su – ” in docker container with error:
    su: System error

    detailed log:
    May 18 07:53:44 i-ikneuftx su[10]: PAM audit_log_acct_message() failed: Operation not permitted
    May 18 07:53:44 i-ikneuftx su[10]: pam_authenticate: System error May 18 07:53:44 i-ikneuftx su[10]: FAILED su for root by root
    May 18 07:53:44 i-ikneuftx su[10]: – /dev/pts/9 root:root

    mysql 5.6 will start failed because there’s “su – mysql” in script: /etc/init.d/mysql

Leave a Reply