Skip to end of metadata
Go to start of metadata

You are viewing an old version of this page. View the current version.

Compare with Current View Page History

« Previous Version 11 Next »

Introduction

  • File Watching is used to automatically start workflows in the event of a file arrival in a directory.
    • Agents watch directories for files and add a file order for each incoming file.
    • File orders are assigned workflows based on configurable rules.
  • File Watching is a flexible means for enterprise application integration (EAI):
    • It enables e.g. 3rd-party applications to launch workflows simply by creating files.
    • It allows to trigger workflow execution on a number of Agents based on arrival of a file.

Feature Video

This video explains file watching with JS7.

Trigger Files and Data Files

Trigger Files are typically zero-byte files that are created to trigger execution of a workflow. A workflow can remove such files at any point in time.

Data Files are processed by jobs in a workflow, for example to import reporting data into a data warehouse. Such files are (re-) moved only after processing by their respective jobs.

For both Trigger Files and Data Files the workflow is responsible for removing an incoming file before completion of the workflow.

File Order Sources for Workflows

The File Order Source is a configuration object that holds the directory and file pattern to be watched for by an Agent and that assigns resulting file orders to a workflow.

  • The Order ID of file orders for example is created like this: #<ISO date>#F<Seconds since Jan 1st 1970>-<ID of file order source>:<file name>, where:
    • #<ISO date># is the date of appearance of the file, e.g. 2021-03-17., enclosed by #. The date is calculated for the time zone assigned the File Order Source.
    • F is a qualifier to indicate a File Order Source.
    • <Seconds since Jan 1st 1970>-  is what it says and is followed by a hyphen.
    • <File Order source ID>: is the unique identifier of the File Order Source configuration object followed by a colon.
    • <file name> is the name of the incoming file.
  • Such file orders can be considered triggers for workflow execution, which implies that the workflow can be executed by the monitoring Agent or by any other Agent.

File Order Sources are managed in the Configuration -> Inventory view like this:


Explanation:

  • Workflow Name: A workflow is assigned to which an order is added per incoming file.
  • Agent: An Agent is assigned that performs file watching. Standalone Agents and Agent Clusters can be assigned. In an Agent Cluster the active Director Agent performs file watching, Subagents do not perform file watching.
  • Directory: The directory that the Agent watches for incoming files:
    • Unix: A path can be specified such as /tmp/incoming
    • Windows: A path can be specified with backslashes or forward slashes such as C:\tmp\incoming and C:/tmp/incoming
    • Unix, Windows: OS environment variables can be used that are known to the Agent, for example from its Instance Start Script. Environment variables and constant strings can be concatenated using the ++ operator and considering quoting for constant strings like this:
      • Unix: env("HOME") ++ '/incoming'
      • Windows: env("TMP") ++ '/incoming'
  • Pattern: The pattern to match an incoming file is not a wildcard expression such as *.csv, instead it represents a Java Regular Expression. The pattern has to match the path of an incoming file including the directory hierarchy, not just the file name. Consider the following examples:
    • match any files: .*
    • match files with a .csv extension: .*\.csv$
    • match files that end with a date in yyyy-mm-dd format: .*\d{4}-\d{2}-\d{2}$
  • Delay: The delay in seconds for which a file is checked to be stable and does not change its size or timestamp.

File Handling

Checking for Incoming Files

Timing

The JS7 Agent applies the following procedure for the timely acceptance of incoming files:

  • For Linux environments the Inotify interface is used. This notifies the JS7 Agent in near real-time about incoming files.
  • For other Unix environments including MacOS, AIX etc. the Agent performs polling every 2s based on the implementation provided with the relevant Java Virtual Machine.
  • For Windows environments the relevant API is used, and notifies the JS7 Agent in near real-time.

Steady State Check

When an incoming file arrives in a directory then the file might not have been completely written at the point in time of appearance.

  • In Windows environments, files that are being written by a process cannot be accessed by jobs.
  • In Unix environments, parallel read and write operations to files by jobs and processes are possible - but not desired, as results are unpredictable.

The best way for a client application to handle this situation is to make a file appear in an atomic transaction, for example to use a temporary file name and to rename the file after it has been completely written. Agents trigger files by patterns and therefore will become aware of a file only when the file path matches the pattern.

Should it not be possible for a client process that writes a file to rename the file after completion then the Agent implements a delay to check the steady state of an incoming file.

  • An interval (default: 2 seconds) can be specified. The Agent will wait for this time and then check if the time stamp or the size of the file changed.
  • If the file has changed then the Agent will wait for the next interval to before applying the same check.

Ghost Appearance

If a file appears then the Agent will create a file order. If the file disappears later on then this has no impact on the file order. However, if a file with the same name appears once again while the current file order is in progress then the Agent will prepare a new file order that will be visible only after the initial file order has completed the workflow.

It is not considered good practice for a client application to make files with the same name appear and disappear in short sequence. However, the situation will be handled by the Agent that creates additional file orders.

Behavior with Outages

If an Agent was not available when a file has been added to a directory then after start of the Agent it will pick up the file and will create a file order.

If the Controller is not available at the point in time when a file order is created then the Agent:

  • can process a workflow for the file order provided that all jobs in this workflow are assigned to this Agent,
  • has to wait for the Controller to become available in order to forward the file order.

Moving Files and Removing Files on Workflow Completion

JS7 does not move or remove files, it's the workflow's responsibility to guarantee that on completion of a workflow the incoming file is no longer present.

  • A job in a workflow can:
    • move an incoming file to an archive location that is not subject to file watching,
    • remove an incoming file.

If an incoming file is still present on completion of the workflow then the file order is moved to a COMPLETED state.

  • The COMPLETED state does not indicate that previous jobs in the workflow were not successful. It can indicates that the incoming file still is available having passed all nodes of the workflow.
  • Users have to make completed file orders leave the workflow and move the remaining file to some other location.



  • No labels