JuiCI is a CI server that has a notion of queuing and priority.
It's designed to work well with agent99 but will play nicely with most frontends to CI.
- Callbacks are created as builds are requested
- Builds are executed in parallel, with one concurrent build per workspace
- Queues can be dynamically created
- Build status visualised
If you create child processes in modules/plugins then you need to register your disinterest or JuiCI will think they're builds and that would be bad.
JuiCI is deliberately very light on the setup front.
bundle install
bundle exec bin/juiciis all you need to have a working instance (provided that you have mongo installed).
Make sure you don't do something innocuous like
bundle install --path .bundlethis might look sane (and it is, kinda) but owing to a quirk in bundler, it will break any ruby code you try to build.
I'm working on a workaround, but in the meantime the fix is to not do it!
JuiCI is very focused on minimal configuration; meaning that beyond starting the server and pointing it at a MongoDB instance, you do not need to do anything special to build a new project. Just request a build; however this means that on your first build you will need to send the commands to create your test environment)
There is a sample implementation of a JuiCI client bundled with the JuiCI source
code, or exposed in the juici-interface gem.
Example:
juicic build --host $hostname --command - --title "test build" \
--project "some project" <<EOF
if [ ! -d .git ]; then
git init .
git remote add origin git://github.com/richo/twat.git
fi
git fetch
git checkout -fq origin/master
set -e
bundle install
bundle exec rake spec"
EOFUsing a convention like script/cibuild as in janky/hubot etc is advisable,
although bear in mind that the logic to checkout the repo will need to be
seperate.
JuiCI supports the notion of priority. Builds given without a priority will be assigned priority 1 (to allow for marking a build as less important with priority 0).
If juici receives a new build with priority higher than any currently unfinished, it will pause whatever it's doing and build the new project. If there is a tie for priority, a FIFO queue is assumed.
JuiCI uses SIGSTOP and SIGCONT internally for job control.
You may specify one or more callbacks when you request a build. They will be called with an (as yet unformalised) json body as the body if/when the build reaches that state. Alternately you may specify "any" as the callback state and it will be called on all state changes.
Apps written in ruby wanting to interact with Juici can include the
juici-interface gem, which presently exposes a few constants to line up with
JuiCI's internal state.
Over time this will be expanded, but for now they are:
Juici::BuildStatus::PASS
Juici::BuildStatus::FAIL
Juici::BuildStatus::START
Juici::BuildStatus::WAITJuiCI poses some interesting security concerns. First off, it will allow anyone with access to run arbitrary commands on your server. I have deliberately not implemented any kind of security inside JuiCI, it plays nicely as a Rack application, and middlewares are much better suited to this task.
It should go without saying, but any builds started by JuiCI will inherit its environment. This means that if you run it in dev mode and forward your ssh agent, builds can ssh to other machines as you!
When running in production you should take steps to ensure that the user JuiCI runs as is no more privileged than it needs to be, and sanitise its environment before execution.
JuiCI by default invokes everything in a subshell - indeed this is the only way to approach this if you want to execute more than one command.
What this means to you as the user though is that unless you go to lengths to
specifically implement it, your process won't see any of the signal handling
madness. The shell(/bin/sh) will see everything, and if killed, your
processes will become orphaned, but carry on.
JuiCI's code lives on Github and the author can be contacted on Twitter
(c) Richo Healey 2012, richo@psych0tik.net
Released under the terms of the MIT license.