The worker processes the same files again when I restart it



If you set up a watch folder in the worker node it uses several techniques to decide whether a file has already been processed or not:

  • within the current worker session (and for up to 14 days afterwards) the worker keeps a list of completed tasks in its task list
  • if your task performs a transcode then the worker will normally skip the file if the destination movie already exists (unless you check the “Overwrite” conversion option)
  • if your task imports then publishes clips to the server then in the publish section you can check the “Avoid duplicates” option

The most efficient approach, however, is to set up the watch folder as a drop box. Users can drop new files and folders onto the drop box as they’re ingested, but the files don’t stay there: your watch action will move the files out of the drop box into a permanent location in your library. This way the worker doesn’t need to keep rescanning all your files, plus you can control access to the library. Only the drop box needs to be publicly writable, while the main library location is set to be read only.

A similar principle applies to tasks triggered by a server query. If you have a watch action that looks for clips with a user field of “Build H264 Proxy” say, then you should set it to something different such as “Proxy Completed” afterwards to prevent the server query from repeatedly finding the same clips.