Stanislaw Adaszewski 69a57df269 | 4 anni fa | |
---|---|---|
example | 4 anni fa | |
focker | 4 anni fa | |
ports | 4 anni fa | |
scripts | 4 anni fa | |
tests | 4 anni fa | |
.gitignore | 4 anni fa | |
README | 4 anni fa | |
README.md | 4 anni fa | |
requirements.txt | 4 anni fa | |
setup.py | 4 anni fa |
Focker is a FreeBSD image orchestration tool in the vein of Docker.
In order to use Focker you need a ZFS pool available in your FreeBSD installation.
Run:
pip install focker
Run:
git clone https://github.com/sadaszewski/focker.git
cd focker/
python setup.py install
or (if you want an uninstaller):
git clone https://github.com/sadaszewski/focker.git
cd focker/
python setup.py sdist
pip install dist/focker-0.9.tgz
Upon first execution of the focker
command, Focker will automatically create the necessary directories and ZFS datasets. You just need to exclude the unlikely case that you are already using /focker
in your filesystem hierarchy. The layout after initialization will look the following:
/focker
/focker/images
/focker/jails
/focker/volumes
images
, jails
, and volumes
have corresponding ZFS datasets with canmount=off
so that they serve as mountpoint anchors for child entries.
To bootstrap the images system you need to install FreeBSD in jail mode to a ZFS dataset placed in /focker/images
and provide two user-defined properties - focker:sha256
and focker:tags
. One way to achieve this would be the following (using Bash shell):
TAGS="freebsd-latest freebsd-$(freebsd-version | cut -d'-' -f1)"
VERSION="FreeBSD $(freebsd-version)"
SHA256=$(echo -n ${VERSION} | sha256)
NAME=${SHA256:0:7}
zfs create -o focker:sha256=${SHA256} -o focker:tags="${TAGS}" zroot/focker/images/${NAME}
bsdinstall jail /focker/images/${NAME}
zfs set readonly=on zroot/focker/images/${NAME}
zfs snapshot zroot/focker/images/${NAME}@1
At this point, Focker is ready to use.
focker
command syntaxThe focker
command is the single entrypoint to all of the Focker's functionality. The overview of its syntax is presented below as a tree where the focker
command is the root, the first level of descendants represents the choice of Level 1 mode (image
, jail
, volume
or compose
), the second level - the Level 2 mode (dependent on L1 mode) and the final third level lists required and optional arguments specific to the given combination of L1/L2 modes.
focker
|- image|img|im|i
| |- build|b
| | |- FOCKER_DIR
| | `- --tags|-t TAG [...TAG]
| |- tag|t
| | |- REFERENCE
| | `- TAG [...TAG]
| |- untag|u
| | `- TAG [...TAG]
| |- list|ls|l
| | `- --full-sha256|-f
| |- prune|p
| `- remove|r
| |- REFERENCE
| `- --remove-dependents|-R
|- jail|j
| |- create|c
| | |- IMAGE
| | |- --command|-c COMMAND (default: /bin/sh)
| | |- --env|-e VAR1:VALUE1 [...VARN:VALUEN]
| | |- --mounts|-m FROM1:ON1 [...FROMN:ONN]
| | `- --hostname|-n HOSTNAME
| |- start|s
| | `- REFERENCE
| |- stop|S
| | `- REFERENCE
| |- remove|r
| | `- REFERENCE
| |- exec|e
| | |- REFERENCE
| | `- [...COMMAND]
| |- oneshot|o
| | `- IMAGE
| | `- --env|-e VAR1:VALUE1 [...VARN:VALUEN]
| | `- --mounts|-m FROM1:ON1 [...FROMN:ONN]
| | `- [...COMMAND]
| |- list|ls|l
| | `- --full-sha256|-f
| |- tag|t
| | |- REFERENCE
| | `- TAG [...TAG]
| |- untag|u
| | `- TAG [...TAG]
| `- prune|p
| `- --force|-f
|- volume|vol|v
| |- create
| | `- --tags|-t TAG [...TAG]
| |- prune
| |- list
| | `- --full-sha256|-f
| |- tag
| | |- REFERENCE
| | `- TAG [...TAG]
| `- untag
| `- TAG [...TAG]
`- compose|comp|c
|- build
| `- FILENAME
`- run
|- FILENAME
`- COMMAND
Individual combinations are briefly described below:
The focker image
mode groups commands related to Focker images.
Build a Focker image according to the specification in a Fockerfile present in the specified FOCKER_DIR. Fockerfile syntax is very straightforward and explained below.
Applies one or more tags to the given image. REFERENCE can be the SHA256 of an image or one of its existing tags. It can be just a few first characters as long as they are unambiguous.
Removes one or more image tags.
Lists existing Focker images, optionally with full SHA256 checksums (instead of the default 7 first characters).
Greedily removes existing Focker images without tags and without dependents.
Removes the specified image.
The focker jail
mode groups commands related to Focker-managed jails.
Creates a new Focker-managed jail. A jail consists of a clone of the given IMAGE
and an entry in /etc/jail.conf
. The configuration entry uses exec.prestart
and exec.start
to specify how the runtime environment (mounts and environmental variables) should be set up. It also calls COMMAND
as last in exec.start
. If not specified COMMAND
defaults to /bin/sh
. The hostname can be specified using the HOSTNAME
parameter. Mounts and environment variables are provided as tuples separated by a colon (:). The environmental variable specification consists of variable name followed by variable value. The mount specification consists of the “from path”, followed by the “on path”. “From path” can be a local system path or a volume name.
Starts the given jail specified by REFERENCE
. REFERENCE
can be the SHA256 of an existing jail or one of its existing tags. It can be just a few first characters as long as they are unambiguous. This command is equivalent of calling jail -c
.
Stops the given jail specified by REFERENCE
. This command is equivalent to calling jail -r
.
Removes the given jail specified by REFERENCE
. The jail is stopped if running, any filesystems mounted under its root directory are unmounted, its ZFS dataset and entry in /etc/jail.conf
are removed.
Executes given COMMAND
(or /bin/sh
if not specified) in the given running jail specified by REFERENCE
. This command is the equivalent of calling jexec
.
Create a new one-time Focker-managed jail. The syntax and logic is identical to focker jail create
, the difference being that the hostname cannot be specified and that the jail will be automatically removed when the COMMAND
exits.
Example: focker jail oneshot freebsd-latest -e FOO:bar -- ls -al
Lists Focker-managed jails. For running jails their JIDs will be displayed.
Applies one or more tags to the given jail. REFERENCE can be the SHA256 of a jail or one of its existing tags. It can be just a few first characters as long as they are unambiguous.
Removes one or more jail tags.
Removes existing Focker jails without tags.
The focker volume
mode groups commands related to Focker volumes.
Create a new Focker volume optionally tagged with the given TAG
s.
Removes existing Focker volumes without tags.
Lists existing Focker volumes. Full SHA256 is displayed if the -f
switch is used, otherwise only the first 7 characters will be shown.
Applies one or more tags to the given volume. REFERENCE can be the SHA256 of a volume or one of its existing tags. It can be just a few first characters as long as they are unambiguous.
Removes one or more volume tags.
The focker compose
mode groups commands related to Focker composition files - focker-compose.yml
.
Builds images, volumes and jails according to the specification provided in the file pointed to by FILENAME
.
Runs one of the commands (specified by COMMAND
) from the composition file pointed to by FILENAME
.
Fockerfile
syntaxA sample Fockerfile
is pasted below.
base: freebsd-latest
steps:
- copy:
- [ '/tmp/x', '/etc/x' ]
- [ 'files/y', '/etc/y' ]
- copy: [ files/z, /etc/z ]
- run: |
pkg install -y python3
- run:
- pkg install -y py37-pip
- pkg install -y py37-yaml
- pkg install -y py37-certbot
- run: |
mkdir -p /persist/etc/ssh && \
sed -i '' -e 's/\/etc\/ssh\/ssh_host_/\/persist\/etc\/ssh\/ssh_host_/g' /etc/rc.d/sshd && \
sed -i '' -e 's/\/etc\/ssh\/ssh_host_/\/persist\/etc\/ssh\/ssh_host_/g' /etc/ssh/sshd_config && \
sed -i '' -e 's/#HostKey/HostKey/g' /etc/ssh/sshd_config
Fockerfile
currently supports only two entries - base
and steps
. base
specifies the parent image on top of which the operations described by steps
are executed. Each entry in steps
results in creation of a new image. Focker determines a checksum for each step and if the corresponding image already exists the step is skipped and work continues on top of the existing image. This is a powerful paradigm for image building experimentation where we can split the task into multiple steps and resume work from the last successful step in case of problems. It is a big time saver. steps
is a list that can contain copy
and run
entries. The copy
entry specifies a single one or a list of copy operations from local files to the image in form of the [FROM, TO]
tuples. The run
entry specifies a chain of commands to be executed within the image. It can be a list of string or a single string.
focker-compose.yml
syntaxA sample composition file illustrating all of the principles is pasted below.
images:
wordpress-5: /path/to/wordpress_5_focker_dir
jails:
wordpress:
image: wordpress-5
env:
SITE_NAME: Test site
mounts:
wp-volume2: /mnt/volume2
wp-volume1: /mnt/volume1
ip4.addr: 127.0.1.1
interface: lo1
volumes:
wp-volume1: {}
wp-volume2: {}
wp-backup: {}
commands:
backup:
jail: wordpress
command: |
mysqldump >/mnt/volume2/backup.sql
mounts:
wp-backup: /mnt/backup
restore:
jail: wordpress
command: |
mysql </mnt/volume2/backup.sql
mounts:
wp-backup: /mnt/backup
The images
entry in Focker composition file specifies a dictionary from image tags to Focker directories (directories containing the Fockerfile
and any supplementary files needed to build an image). Upon running focker compose build
Focker will run focker image build
for all of the specified directories and tag the results with the corresponding tags. This process can be repeated without significant performance penalty since the images will not need to be rebuilt unless their Fockerfile
s or contexts change.
The jails
entry in the Focker composition file specifies a dictionary from jail tags to jail specifications. A jail specification is a dictionary that can contain the following fields: image
, env
, mounts
, exec.start
, exec.stop
, hostname
, ip4.addr
, interface
. image
, env
and mounts
have the same semantics as in the focker jail create
command. The syntax for env
and mounts
is in the form of dictionaries. exec.start
, exec.stop
, hostname
, ip4.addr
and interface
have the same semantics as the corresponding entries in /etc/jail.conf
. The jails will be recreated each time focker compose build
is executed. Hence, any persistent data should be stored in volumes.
The volumes
entry in the Focker composition file specifies a dictionary from volume tags to volume specifications. Currently a volume specification must be an empty dictionary. Specified volumes will be created by focker compose build
and tagged with corresponding tags unless volumes with given tags already exist. Volumes are meant to persist data beyond the jail lifecycle.
The commands
entry in the Focker composition file specifies a dictionary from command names to command specifications. A command specification can contain the following fields: jail
, mounts
and command
. The jail
field specifies in which jail the given command should be executed (the jail must be already running). The mounts
entry specifies additional mounts that will be used only during the execution of the command. Finally the command
entry specifies the command itself using the same syntax as the run
step in a Fockerfile
.
The best way to learn is by practice. Take a look at the example and start writing your own Focker specifications.
Focker provides powerful containerization primitives (images, volumes and containers) first introduced by the Docker platform without taking up the significantly more challenging task of achieving Docker compatibility. This has never been and never will be the goal of Focker which allows it to remain a lightweight tool with minimal dependencies and highly maintainable codebase. At the same time, the image building paradigm based on checksummed steps/layers and flexible composition builder offer significant time savings to pragmatic sysadmins.