Guide to the Allura task and event system¶
Our event system is driven by a MongoDB-based queuing system, most of which you can ignore, because we’ve simplified it down to two ideas: tasks and event handlers.
Before we get into the details perhaps a few definitions are in order:
app – tool for allura such as the tracker, scm, or wiki apps
task – callable defined in a app that gets invoked by the taskd daemon
event handler – callable defined in an app that gets called on message events. Event handlers are identified by a string name, so a single event can fan out to multiple callables.
The MonQTask class is central to the Allura asynchronous processing system. Simply put, a MonQTask is a document in MongoDB that contains a context (project/app/user), a function pointer (specified as a dotted string), and arguments to that function. Tasks are scheduled by creating MonQTask documents, and the taskd daemon executes them as though they were happening in the web context.
To simplify the use of tasks, Allura provides a decorator @task that marks a function as ‘taskable.’ This decorator adds a .post method to the function object that allows the function to be scheduled as a MonQTask. For instance, the commit task (flushing Solr caches) is defined in as the following:
In order to schedule this task for execution by taskd, simply use the .post method:
If we wanted to call commit directly (e.g. in a test), we can still do that as well:
Events provide fanout capability for messages, letting several functions get called in response to the same ‘event.’ To note a function as an event handler, you use the @event_handler decorator. For instance, there is an event handler on all project updates to subscribe the project’s admins to project changes:
In order to invoke all the event handlers for a particular topic, we use the g.post_event helper:
Under the covers, this is scheduling an event task that calls all the handlers for a particular named event. Note that you can pass arguments (*args, and **kwargs) to event handlers just like you do to tasks, with the exception that the topic name (above, this would be ‘project_updated’) is always the first parameter passed to the event handler.
Running the Task Daemon¶
In order to actually run the asynchronous tasks, we have written a paster command taskd. This creates a configurable number of worker processes that watch for changes to the MonQTask collection and execute requested tasks. taskd can be run on any server, but should have similar access to the MongoDB databases and configuration files used to run the web app server, as it tries to replicate the request context as closely as possible when running tasks.