Plone on Heroku
** WARNING ** Currently Plone contains syntax errors (skins aren't valid python, why are we calling them .py!!) and heroku really doesn't like that. Currently you can't deploy again after the first successful Plone deployment because of it. I have reported this.
You will require
- The heroku tools installed locally and working
- A verified heroku account
- A willingness to spend $15 a month for 20GB of Postgres (Plone doesnt fit in 5mb).
Starting a project
We need to be able to install Plone in a virtualenv and do without any of the environment building capabilities of buildout. On top of that, the runtime environment is temporary, so we cant build it manually after the initial push.
In your requirements.txt:
isotoma.depends.zope2_13_8 isotoma.depends.plone4_1 isotoma.plone.heroku
The first 2 eggs are virtual packages that install 200+ packages in your virtualenv. These are needed because we need to pin the versions of Plone that are to be used.
The third egg provides
bin/migrate. These helpers
will dynamically provision their environment as it is required and instance
a site / run migrations.
We'll be using RelStorage to get cheap persistent storage. We have to use 1.4.2 over the 1.5.x series in order to avoid the plpgsql requirement it would introduce.
At the time of writing it appears that only Django Python apps get a database automatically. You can get one by adding a folder with an empty settings.py:
You will need a Procfile so heroku knows how to run a zope instance:
web: ./bin/plone -p $PORT
Doing a local build
Build your virtualenv:
virtualenv . ./bin/pip install -r requirements.txt
You can start a plone instance like this:
That will get you a plone instance running on port 8080. By default it won't be using any data store.
Adding your product
You can quickly create a new product with ZopeSkel (version 2.21.2, as the 3.x line is currently broken :D):
source bin/activate mkdir src && cd src pip install zopeskel==2.21.2 zopeskel plone my.app
Then follow the prompts. Make sure to answer yes to the GenericSetup question.
You can add your own custom eggs to requirements.txt:
The instructions tell you to use
file: prefixes. They lie. Don't. You can't
-e because heroku's build environment will hardcode a path it shouldn't.
Your ZCML should be found by z3c.autoinclude.
Then you should be able to install your products throught the ZMI or by using the migrate script described below.
Deploying to heroku
Initialise a git repository (heroku loves git). In the root of your project:
git init .
And then create a file called .gitignore with the following contents:
distribute-0.6.10.tar.gz lib/ bin/ include/ zope/ README.txt
git add . git commit -m "Initial commit" ~/bin/heroku create --stack cedar
This will give you your app name (of the format word-word-9999). Log on to your heroku account and upgrade the shared database from 5MB to 20GB (for $15 a month). And then you can finally deploy:
git push heroku master
Then wait. It should just work, if it doesn't its probably a timeout. It takes a long time to deploy 200+ eggs and heroku probably thinks your deployment has gone wrong and times out.
So edit your requirements:
isotoma.depends.zope2_13_8 # isotoma.depends.plone4_1 isotoma.plone.heroku # src/my.app
Commit and push to heroku.
That will build Zope without Plone, about half of the eggs that needed to be built. Then you can uncomment the Plone dependency egg and push again to finish off.
You should now have a working Plonesite!
Re-rooting your portal
By default your actual site won't be at
/ it will be at
/Plone. We can fix
that with some old school Zope magic - note that the
migrate command can set
this up for you automatically.
In the ZMI, in
/Plonecreate a SiteRoot object. Default settings are fine.
In the ZMI, in
/create a DTMLMethod containing:<dtml-let stack="REQUEST['TraversalRequestNameStack']"> <dtml-if "stack and stack[-1]=='zmi'"> <dtml-call "stack.pop()"> <dtml-call "REQUEST.setVirtualRoot('zmi')"> <dtml-else> <dtml-call "stack.append('Plone')"> </dtml-if> </dtml-let>
In the ZMI, at
/create an AccessRule and point it at the DTMLMethod we just created.
Now any requests for
/foo will be handled by
/Plone/foo and any requests
/zmi/manage will be handled by
The migrate tool
The migrate script uses the plone setup features of
automate setup of your site. It can apply profiles, install products, set properties
and even call random mutators.
Add a migrate.cfg to the root of your project:
[main] # The id of the Plone Site that is created. Default is Plone. site-id = Plone # The admin user that was created by mkzopeinstance - for us it will almost # certainly be admin. Default is admin. admin-user = admin # Whether or not to apply the SiteRoot/AccessRule policy described in the previous # section. Default is False. rootify = True # List of products to install on the initial migrate (when ``/Plone`` is created) products-initial = Products.foo # List of products to install (or reinstall) every time migrate is run products = Products.LinguaPlone # List of GenericSetup profiles to apply the first time migrate is run (when # ``/Plone`` is created) profiles-initial = myapp.policy:initial # List of GenericSetup profiles to apply every time ``bin/migrate`` is run profiles = myapp.policy:default
That one doesn't make any sense, but does show what you can do. To run it locally:
./bin/migrate -c migrate.cfg
And to run against your heroku app:
~/bin/heroku run ./bin/migrate -c migrate.cfg
The default is to look for a migrate.cfg in the root of the branch you you don't
have to tell it that - you can just
Maintaining your app
You can look at your looks with the built in heroku log tool:
You can get an interpreter pointed at your database with the debug command:
./bin/heroku run ./bin/debug
You can run a script in your git repo using the run command:
./bin/heroku run ./bin/run scripts/myscript.py