Bases: pipelet.tracker.Tracker, threading.Thread
Fast sqlite-based tracker.
Tasks are saved in a sqlite3 data base. Format is: TODO
Add a list of task to the queue.
Add a task to the data base, and return it with its new id and status.
Remove queued and running tasks from data base.
This is called only at the init of the scheduler.
Empty the sql requests queue in one go.
As most of the asynchronous requests are updates, and the change are commited to the sqlfile after completion, calling flush too often with few requests may seriously hinder the performances.
Start the tracker.
Use the shared connection to the sqlite data base, and regularly call flush to empty the requests queue, and commit the result to the db.
Track the submission of objects through files in the segment directory.
The numerous file access may result in a serious performance bottleneck. Nevertheless, it is the only work tracker that may be compatible with the ancestral version of pipelet.