Class Worker

source code

Worker is an essential base class for the ClusterShell library. The goal of a worker object is to execute a common work on a single or several targets (abstract notion) in parallel. Concret targets and also the notion of local or distant targets are managed by Worker's subclasses (for example, see the DistantWorker base class).

A configured Worker object is associated to a specific ClusterShell Task, which can be seen as a single-threaded Worker supervisor. Indeed, the work to be done is executed in parallel depending on other Workers and Task's current paramaters, like current fanout value.

ClusterShell is designed to write event-driven applications, and the Worker class is key here as Worker objects are passed as parameter of most event handlers (see the ClusterShell.Event.EventHandler class).

The following public object variables are defined on some events, so you may find them useful in event handlers:

• worker.current_node [ev_pickup,ev_read,ev_error,ev_hup] node/key concerned by event
• worker.current_msg [ev_read] message just read (from stdout)
• worker.current_errmsg [ev_error] error message just read (from stderr)
• worker.current_rc [ev_hup] return code just received

Example of use:

>>> from ClusterShell.Event import EventHandler
>>> class MyOutputHandler(EventHandler):
...     def ev_read(self, worker):
...             node = worker.current_node
...             line = worker.current_msg
...             print "%s: %s" % (node, line)
...

Read worker stream buffer.

Return stream read buffer of current worker.

Arguments:
    node -- node name; can also be set to None for simple worker
            having worker.key defined (default is None)
    sname -- stream name (default is 'stdout')