The Sali lab web framework aims to provide a simple set of classes and modules to simplify the process of deploying a web service. It is designed for typical Sali lab web applications - that is to say, jobs that are submitted from a web interface but that then run on one or more cluster nodes for a possibly long period of time.
Developers eager to get started may want to jump straight to Quick start.
The framework is split up into four distinct parts:
The frontend consists of HTML pages and Python scripts that provide a web interface to interact with an end user. It also provides an XML-based interface to allow automated submission of jobs. It handles the uploading of input files, the initial submission of jobs, displaying a queue of all jobs in the system, and showing the results of completed jobs. It can also potentially handle user logins.
The backend is a set of Python classes that manages jobs after the initial submission. This typically runs as a daemon on our modbase machine, picking up submitted jobs from the frontend, submitting jobs to the cluster and gathering results, and doing any necessary pre- or post-processing.
The build system is a set of extensions to SCons that simplifies the procedure of deploying a a web service (initial setup, and installing everything in the correct locations and with the right permissions).
Monitoring is typically set up by a sysadmin, and ensures that the web service, once set up and made available to end users, is correctly functioning.
The web framework stores all persistent data about the jobs in a MySQL database, and configuration for the service is stored in a set of configuration files that are shared between all parts of the framework.
The framework aims to be:
Comprehensive: can handle simple services like ModLoop and complex services like CM-MR.
Robust: full error checking is done at each step of the process; job failures are caught and reported to the server admin.
Extensible: a service can be set up without user authentication, but this can be simply added in later. The backend Python classes provide multiple hooks to allow each step of the job to be modified.
Secure: by default, file permissions should be set up sensibly. There should be no need for world-writeable directories or other hacks to allow files to be passed from the web server to the cluster. The backend runs as a different user from the web server, so it is unlikely that a bug in the frontend can be used to break into the cluster.
Efficient: it is straightforward to configure the system so that long-term data from jobs are not left on expensive disks such as NetApp, but are instead moved to park disks or deleted.
Easy to use: given a simple workflow, it should be easy to deploy a simple web server that implements it.