This document introduces Docker and provides an overview of its key features and benefits. It explains that Docker allows developers to package applications into lightweight containers that can run on any Linux server. Containers deploy instantly and consistently across environments due to their isolation via namespaces and cgroups. The document also summarizes Docker's architecture including storage drivers, images, and the Dockerfile for building images.
3. @jpetazzo
●
Wrote dotCloud PAAS deployment tools
– EC2, LXC, Puppet, Python, Shell, ØMQ...
●
Docker contributor
– Docker-in-Docker, VPN-in-Docker,
router-in-Docker... CONTAINERIZE ALL THE THINGS!
●
Runs Docker in production
– You shouldn't do it, but here's how anyway!
10. Deploy reliably & consistently
● If it works locally, it will work on the server
● With exactly the same behavior
● Regardless of versions
● Regardless of distros
● Regardless of dependencies
11. Deploy efficiently
● Containers are lightweight
– Typical laptop runs 10-100 containers easily
– Typical server can run 100-1000 containers
● Containers can run at native speeds
– Lies, damn lies, and other benchmarks:
http://qiita.com/syoyo/items/bea48de8d7c6d8c73435
13. Is there really
no overhead at all?
● Processes are isolated,
but run straight on the host
● CPU performance
= native performance
● Memory performance
= a few % shaved off for (optional) accounting
● Network performance
= small overhead; can be reduced to zero
15. High level approach:
it's a lightweight VM
● Own process space
● Own network interface
● Can run stuff as root
● Can have its own /sbin/init
(different from the host)
« Machine Container »
16. Low level approach:
it's chroot on steroids
● Can also not have its own /sbin/init
● Container = isolated process(es)
● Share kernel with host
● No device emulation (neither HVM nor PV)
« Application Container »
17. How does it work?
Isolation with namespaces
● pid
● mnt
● net
● uts
● ipc
● user
18. pid namespace
jpetazzo@tarrasque:~$ ps aux | wc -l
212
jpetazzo@tarrasque:~$ sudo docker run -t -i ubuntu bash
root@ea319b8ac416:/# ps aux
USER PID %CPU %MEM VSZ RSS TTY STAT START TIME COMMAND
root 1 0.0 0.0 18044 1956 ? S 02:54 0:00 bash
root 16 0.0 0.0 15276 1136 ? R+ 02:55 0:00 ps aux
(That's 2 processes)
23. user namespace
● No demo, but see LXC 1.0 (just released)
● UID 0→1999 in container C1 is mapped to
UID 10000→11999 in host;
UID 0→1999 in container C2 is mapped to
UID 12000→13999 in host; etc.
● what will happen with copy-on-write?
– double translation at VFS?
– single root UID on read-only FS?
24. How does it work?
Isolation with cgroups
● memory
● cpu
● blkio
● devices
25. memory cgroup
● Keeps track pages used by each group:
– file (read/write/mmap from block devices; swap)
– anonymous (stack, heap, anonymous mmap)
– active (recently accessed)
– inactive (candidate for eviction)
● Each page is « charged » to a group
● Pages can be shared (e.g. if you use any COW FS)
● Individual (per-cgroup) limits and out-of-memory killer
26. cpu and cpuset cgroups
● Keep track of user/system CPU time
● Set relative weight per group
● Pin groups to specific CPU(s)
– Can be used to « reserve » CPUs for some apps
– This is also relevant for big NUMA systems
27. blkio cgroups
● Keep track IOs for each block device
– read vs write; sync vs async
● Set relative weights
● Set throttle (limits) for each block device
– read vs write; bytes/sec vs operations/sec
Note: earlier versions (<3.8) didn't account async correctly.
3.8 is better, but use 3.10 for best results.
28. devices cgroups
● Controls read/write/mknod permissions
● Typically:
– allow: /dev/{tty,zero,random,null}...
– deny: everything else
– maybe: /dev/net/tun, /dev/fuse, /dev/kvm, /dev/dri...
● Fine-grained control for GPU, virtualization, etc.
29. How does it work?
Copy-on-write storage
● Create a new machine instantly
(Instead of copying its whole filesystem)
● Storage keeps track of what has changed
● Since 0.7, Docker has a storage plugin system
30. Storage:
many options!
Union
Filesystems
Snapshotting
Filesystems
Copy-on-write
block devices
Provisioning Superfast
Supercheap
Fast
Cheap
Fast
Cheap
Changing
small files
Superfast
Supercheap
Fast
Cheap
Fast
Costly
Changing
large files
Slow (first time)
Inefficient (copy-up!)
Fast
Cheap
Fast
Cheap
Diffing Superfast Superfast Slow
Memory usage Efficient Efficient Inefficient
(at high densities)
Drawbacks Random quirks
AUFS not mainline
!AUFS more quirks
ZFS not mainline
BTRFS not as nice
Higher disk usage
Great performance
(except diffing)
Bottom line Ideal for PAAS and
high density things
This is the Future
(probably)
Dodge Ram 3500
45. Docker-what?
The Big Picture
● Open Source engine to commoditize LXC
● Using copy-on-write for quick provisioning
● Allowing to create and share images
● Standard format for containers
(stack of layers; 1 layer = tarball+metadata)
● Standard, reproducible way to easily build
trusted images (Dockerfile, Stackbrew...)
46. Docker-what?
History
● Rewrite of dotCloud internal container engine
– original version: Python, tied to dotCloud PaaS
– released version: Go, legacy-free
47. Docker-what?
Under the hood
● The Docker daemon runs in the background
– manages containers, images, and builds
– HTTP API (over UNIX or TCP socket)
– embedded CLI talking to the API
48. Docker-what?
Take me to your dealer
● Open Source
– GitHub public repository + issue tracking
https://github.com/dotcloud/docker
● Nothing up the sleeve
– public mailing lists (docker-user, docker-dev)
– IRC channels (Freenode: #docker #docker-dev)
– public decision process
49. Docker-what?
The ecosystem
● Docker Inc. (formerly dotCloud Inc.)
– ~30 employees, VC-backed
– SAAS and support offering around Docker
● Docker, the community
– more than 360 contributors, 1600 forks on GitHub
– dozens of projects around/on top of Docker
– x100k trained developers
50. One-time setup
● On your servers (Linux)
– Packages (Ubuntu, Debian, Fedora, Gentoo, Arch...)
– Single binary install (Golang FTW!)
– Easy provisioning on Rackspace, Digital Ocean, EC2, GCE...
● On your dev env (Linux, OS X, Windows)
– Vagrantfile
– boot2docker (25 MB VM image)
– Natively (if you run Linux)
51. The Docker workflow 1/2
● Work in dev environment
(local machine or container)
● Other services (databases etc.) in containers
(and behave just like the real thing!)
● Whenever you want to test « for real »:
– Build in seconds
– Run instantly
52. The Docker workflow 2/2
Satisfied with your local build?
● Push it to a registry (public or private)
● Run it (automatically!) in CI/CD
● Run it in production
● Happiness!
Something goes wrong? Rollback painlessly!
54. 1) docker run ubuntu bash
2) apt-get install this and that
3) docker commit <containerid> <imagename>
4) docker run <imagename> bash
5) git clone git://.../mycode
6) pip install -r requirements.txt
7) docker commit <containerid> <imagename>
8) repeat steps 4-7 as necessary
9) docker tag <imagename> <user/image>
10) docker push <user/image>
56. Authoring images
with run/commit
● Pros
– Convenient, nothing to learn
– Can roll back/forward if needed
● Cons
– Manual process
– Iterative changes stack up
– Full rebuilds are boring, error-prone
58. FROM ubuntu
RUN apt-get -y update
RUN apt-get install -y g++
RUN apt-get install -y erlang-dev erlang-manpages erlang-base-hipe ...
RUN apt-get install -y libmozjs185-dev libicu-dev libtool ...
RUN apt-get install -y make wget
RUN wget http://.../apache-couchdb-1.3.1.tar.gz | tar -C /tmp -zxf-
RUN cd /tmp/apache-couchdb-* && ./configure && make install
RUN printf "[httpd]nport = 8101nbind_address = 0.0.0.0" >
/usr/local/etc/couchdb/local.d/docker.ini
EXPOSE 8101
CMD ["/usr/local/bin/couchdb"]
docker build -t jpetazzo/couchdb .
60. Authoring images
with a Dockerfile
● Minimal learning curve
● Rebuilds are easy
● Caching system makes rebuilds faster
● Single file to define the whole environment!
66. Stability and performance
● Many, many, many bugfixes
● Performance improvements
– When Docker starts
– When creating/destroying containers
(Especially en masse)
– Better memory footprint
67. ADD caching
● ADD no longer breaks caching
● You can now use the following pattern:
ADD requirements.txt /src/requirements.txt
RUN pip install r requirements.txt
ADD . /src
68. New ONBUILD instruction
● Register « triggers » to be executed later,
when building an image on top of this one
● Triggers are executed in downstream context
Example:
RUN aptget install buildessential
ONBUILD ADD . /src
ONBUILD RUN cd /src; ./configure; make install
70. Storage:
many options!
Union
Filesystems
Snapshotting
Filesystems
Copy-on-write
block devices
Provisioning Superfast
Supercheap
Fast
Cheap
Fast
Cheap
Changing
small files
Superfast
Supercheap
Fast
Cheap
Fast
Costly
Changing
large files
Slow (first time)
Inefficient (copy-up!)
Fast
Cheap
Fast
Cheap
Diffing Superfast Superfast Slow
Memory usage Efficient Efficient Inefficient
(at high densities)
Drawbacks Random quirks
AUFS not mainline
!AUFS more quirks
ZFS not mainline
BTRFS not as nice
Higher disk usage
Great performance
(except diffing)
Bottom line Ideal for PAAS and
high density things
This is the Future
(probably)
Dodge Ram 3500
71. BTRFS storage driver
● Available on « normal » kernels
● Supposedly faster than DM (but YMMV)
● Doesn't use BTRFS delta stream yet
(i.e. commit/diff is not optimized yet)
● Use it, abuse it, break it!
72. Socket activation
● Plays nice with systemd
● Avoids race condition in boot scripts
Docker doesn't handle API requests when it's
starting, so without socket activation, you need
e.g. reconnection logic.
73. OS X support
● The CLI now runs on OS X
● boot2docker is awesome
79. Coming Soon
● Network acceleration
● Container-specific metrics
● Plugins (e.g. for logging)
● Orchestration hooks
Those things are already possible,
but will soon be part of the core.