The imaginator daemon builds OS and OS+application images which are specified by image manifests.
The imaginator provides a web interface on port
6975
which shows a status page, links to built-in
dashboards and access to performance metrics and logs. If
imaginator is running on host myhost
then the URL
of the main status page is http://myhost:6975/
. An RPC over
HTTP interface is also provided over the same port.
Imaginator is started at boot time, usually by one of the provided init scripts. The imaginator process is baby-sat by the init script; if the process dies the init script will re-start it. It may be stopped with the command:
service imaginator stop
which also kills the baby-sitting init script. It may be started with the command:
service imaginator start
There are many command-line flags which may change the behaviour of imaginator, generally with defaults which should be adequate for most deployments. Built-in help is available with the command:
imaginator -h
The init script reads configuration parameters from the
/etc/default/imaginator
file. The following is the minimum
likely set of parameters that will need to be configured.
The CONFIGURATION_URL
variable specifies the main
configuration URL, which may be a local file or a HTTP URL. It is
recommended to specify a HTTP URL which references a file in a Git
repository.
The IMAGE_SERVER_HOSTNAME
variable specifies the
hostname where the imageserver is running. This
hostname must be resolvable by the imaginator.
The VARIABLES_FILE
variable specifies an optional
filename from which to read special variables which may be used for
variable expansion. These are typically used to store secrets for
accessing Git repositories which require authentication. Each line
should contain a single NAME=Value
entry.
RPC access is restricted using TLS client authentication.
Imaginator expects a root certificate in the file
/etc/ssl/CA.pem
which it trusts to sign certificates which
grant access. It also requires a certificate and key which grant it the
ability to add and get images and objects from the imageserver.
These should be in the files /etc/ssl/imaginator/cert.pem
and /etc/ssl/imaginator/key.pem
, respectively.
The builder-tool utility may be used to request the imaginator to build an image.
The main configuration URL points to a JSON encoded file that describes all the image streams and how to build them. The top-level JSON object defines the following fields:
BindMounts
: a list of directories that will be
bind-mounted into the build environmentsBootstrapStreams
: a table of bootstrap image
stream names and their respective configurationsImageStreamsCheckInterval
: the interval between checks
for updated image streamsImageStreamsToAutoRebuild
: an array of image
stream names that should be rebuilt periodically, in addition to
bootstrap streams that are always rebuilt automaticallyImageStreamsUrl
: the URL of a configuration file
containing a list of all the user-defined image streamsMtimesCopyFilterLines
: an array of regular expressions
matching files which should be excluded from copying mtimes from older
images (when only the mtime is different). Python files should probably
be specified herePackagerTypes
: a table of packager type names
(i.e. deb
and rpm
) and their respective
configurationsRelationshipsQuickLinks
: a list of
Name
,URL
tuples to display on the image
streams relationships dashboard. Useful for customisationA sample configuration file is provided which
may be modified to suit your environment. This is a fully working
configuration and only requires modification of the location of the
package repositories and the ImageStreamsUrl
for your
custom image streams. Details on some of the above
configuration entries are described below.
Each bootstrap stream is configured by a JSON object with the following fields:
BootstrapCommand
: an array of strings containing the
bootstrap script to run to generate the image contents (typically
debootstrap
and yumbootstrap
). The
$dir
variable expands to the root directory of the image to
buildFilterLines
: an array of regular expressions matching
files which should not be included in the imageImageFilterUrl
: a URL from which a filter lines can be
read. The filter will be attached to the imageImageTagsUrl
: a URL from which JSON-encoded key:value
tags can be read. The tags will be attached to the imageImageTriggersUrl
: a URL from which JSON-encoded
triggers can be read. The triggers will be attached to the imagePackagerType
: the name of the packager type to useThis is a JSON encoded configuration file listing all the user-defined image streams. It contains the following top-level fields:
StreamPatterns
: this contains a table of regular
expressions matching image stream names and their respective
configurationsStreams
: this contains a table of image stream
names and their respective configurationsThe configuration for an image stream is a JSON object with the following fields:
BuilderGroups
: a list of groups. Members of these
groups are permitted to build images for this streamBuilderUsers
: a list of users who are permitted to
build images for this streamManifestUrl
: the URL of a Git repository containing the
image
manifest for the image. The special URL scheme dir
points to a local directory tree. Variables specified in the
VARIABLES_FILE
will be expanded hereManifestDirectory
: the directory within the Git
repository containing the image manifest for
the image. If unspecified, the top-level directory in the repository is
used. The $IMAGE_STREAM
variable expands to the name of the
image streamVariables
: key:value variables which may be referred to
in the manifest
file. The values undergo variable
expansionAn example configuration file is provided. Note the use of variables in different places.
Each packager type is configured by a JSON object with the following fields:
CleanCommand
: an array of strings containing the
command to run when cleaning up packager debrisCleanPatterns
: an array of regular expressions which
match pathnames that should be removed when cleaning up packager
debrisInstallCommand
: an array of strings containing the
command to run when installing packages. The name of the package to be
installed is appended to the command-lineListCommand
: a JSON object defining how to list
packages which are installed. This JSON object contains the following
fields:
ArgList
: an array of strings containing the command to
run when listing installed packagesSizeMultiplier
: an optional multiplier to apply to the
output of the listing command to convert the size result to BytesUpdateCommand
: an array of strings containing the
command to run when updating the package databaseUpgradeCommand
: an array of strings containing the
command to run when upgrading the already installed packagesVerbatim
: an array of strings containing commands to
run before any of the above defined commandsThese parameters are used to generate a
/bin/generic-packager
script which is used as an interface
to the native OS packaging tools.