Process Monitoring

Requirements

The user wants to see:

  1. See all the active processes in the local network
    • See the current and past status of all active processes
    • where each process step was executed
  2. See all the deployed processes of one Engine => Machine View
  3. See all the deployed processes on the complete local network
  4. See the status of all “my”, deployed processes (even not in the local network)
    • my mean: all processes that the User is assigned to
    • See all different running versions of a process

Main Problems

  1. How to Display the Status in the MS GUI
  2. How to Get the complete Process (for static and dynamic deployment)
  3. How to Get the Status

Display the Status in the MS GUI

Which information is needed and Why

  • BPMN Definition ID: to display the correct BPMN diagram
  • (Process Version: to display the correct BPMN)
  • Process State: running|ended|stopped|paused|uncertain
  • Instance and Token Info:
    • Amount of running tokens: to visualize the running tokens in the process
    • Flow Node where a token currently is: to visualize a token at a certain place in the BPMN diagram
    • Instance Start Time: to show an overall status when the instance started
    • (Instance End Time: to show an overall status when the instance ended; maybe to show an execution history in the MS)
  • Information for every Flow Node (Activity/Gateway/Event):
    • Machine ID/IP/Name to show on which Machine the Flow Node was executed
    • (Next Machine: to track the execution (this only needs to be shown, if this next machine is not reachable))
    • Flow Node Start Time and End Time: to calculate the Flow Node and the overall Process progress
    • Current progress: if the activity is not yet completed, this is an indicator to show that it is already 30% finished.
  • Current Process Data: to monitor/debug to current process status (problem: can be big for files)
  • Log:
    • Technical Errors

Monitoring GUI

TODO

How to get the Process

For Req 1. and Req 2.: How to get the complete process?

  • Pull: Ask every reachable Engine for the process
    • Problem: Static deployment allows to only send parts of a process (not yet defined) => need to backtrace all referenced Engines, it is possible that some are not reachable for the requesting MS
    • Synchronize the read processes with the locally stored processes
    • PROBLEM: process steps can be outside the current network
  • for static assignment?:
  • for dynamic assignment?:

For Req. 4.: How to get the complete process?

  1. Already locally there, because it’s yours
  2. Sync from Server
    1. Processes from you
    2. Processes assigned to you

Get The Status

Where to get the Data for the PULL method

Data Where/How Calculation
BPMN Definition ID /process/{file-name}
Process State /process/{file-name}/instance/{id} => instanceState running: if at least one token is in the running state,
ended/stopped/paused: if all tokens are in the ended/stopped/paused state,
uncertain: if the instance is always forwarded and nextMachine is not reachable
Instance: Amount tokens /process/{file-name}/instance/{id} => token[n].tokenId amount of unique token ids for an instance
Instance: Start Time /process/{file-name}/instance/{id} => globalStartTime no calculation necessary
Instance: End Time /process/{file-name}/instance/{id} => TODO
Token: Current Flow Node /process/{file-name}/instance/{id} => token[n].currentFlowNodeId no calculation necessary
Flow Node: ID/IP/Name /process/{file-name}/instance/{id} => log.flowNodeExecutions[n].machine Match data with flowNodeId
Flow Node: next Machine /process/{file-name}/instance/{id} => tokens[n].nextMachine get this info from the previous machine
Flow Node: Start Time and End Time /process/{file-name}/instance/{id} => log.flowNodeExecutions[n].startTime/endTime Match data with flowNodeId
Flow Node: Current progress /process/{file-name}/instance/{id} => tokens[n].currentFlowNodeProgress no calculation necessary
Process Data /process/{file-name}/instance/{id} => variables no calculation necessary
Log/Technical Error /logging/process/{file-name}/instance/{id} Display the logs in combination with a Machine ID/IP

To get the state of the complete process with the PULL method, it is necessary to follow the whole chain of Machines which executed a part of the process. The following is done by looking at /process/{file-name}/instance/{id} => tokens[n].nextMachine.

Calculation determines how to calculate the whole process state, and not only displaying the local state of one Machine.

How to get the status info?

  • for static assignment?:
    • 1. Pull: 1. get the status from every engine
      • Problem: trace Engine step by step (+ every Engine needs to be reachable)
    • 2. Push: 1. the Engine sends the process to one or more MS
      • Problem: stop sending, if MS don’t need the status anymore
  • for dynamic assignment?:

    • 1. Pull: 1. get the status from every engine
      • Problem: how to trace the process and Engines step by step (+ every Engine needs to be reachable)
    • 2. Push: 1. the Engine sends the process to one or more MS
      • Problem: stop sending, if MS don’t need the status anymore
  • Show process info of PUSHed processes

    • does need one more available server
    • what happens if one process step does not send any status info (because offline mode)?
  • Synced Server Version: wie werden Status Infos verteilt?

    • direkt an alle MS oder über den Server?
  • Welche Push Methode?

    • MS kann sich direkt registrieren bei Engine (Als wenn Engine eigener Msg Broker ist), entweder für
      • eine Prozess-Instanz => übergeben mit Token an nächste Engine,
      • einen Prozess in einer Version
      • einen Prozess in allen Versionen
      • Alle Prozessen der Engine (?)
      • Problem: De-Registrierung wenn keine Infos mehr haben will => erfordert Abmeldung
    • In Diagramm einen Message Broker hinterlegen: dort Anmelden
    • Wäre auch zum Verfolgen von Prozessen geeignet, die im Netzwerk sind und vielleicht wieder verschwinden (UC 1. und 2.) => Anmeldung am Broker (mit Berechtigung)
  • Wie kann man vorher deployten Prozess nachverfolgen, wenn die Engines dann nicht mehr sichtbar sind, und absichtlich keine Daten senden?

    • anfragen wenn erreichbar
    • absichtlich keine => sollte von der Engine berücksichtigt werden
      • Deshalb kann es passieren, dass einige Aktivitäten einen Status liefern, andere aber aussetzen (alles die selbe Instanz)
  • wie werden IP-Adressen für das statische Deployment gespeichert, so dass beim nächsten Depoy nicht erneut eingegeben werden müssen?

    • mit versioniert
  • Muss ein Log gespeichert werden?

    • Ja, für das dynamische Deployment: welche Aktivität durch welche Maschine ausgeführt wurde
      • Kann entweder durch die Engine im Nachhinein in die Flow Node geschrieben werden (mit machineId and machineAddress) oder in ein Log
      • Problem: Hier wird Status und Prozess vermischt => wie wird das gesendet?
    • Vll, wer alles Status Infos abgegriffen hat
  • Muss die Engine den BPMN-Prozess verändern?

    • Logs für Dynamische Deployment: welches Gerät
    • alle MS, die gerne Status Tokens erhalten wollen

State on one PROCEED Engine

One process execution is usually done on multiple PROCEED Engines. Every Engine manages the instance state of the process execution locally. The question are:

  • which instance states exists on one Engine?

Instance States on a PROCEED Engine

  • The BPMN spec. does not really define a process state diagram, but mentions multiple states (section 10.5.7, page 280): Activated, In execution, Completed, In Compensation, Compensation, In Error, In Cancellation, Cancelled

Our current implementation has the following states, which also results from the implementation state of the used BPMN engine (see wiki): - “instanceState”: “running|paused|forwarded|ended|partly-ended-forwarded|stopped”,

This value is either set with function in the BPMN engine (running, paused, stopped) or calculated by the state of every available token.

Failed state

What does it mean?

  • error not catched

Also technical Error?

Should every other token also be canceled?

  • no, other tasks can run in parallel and we can handle the error manually (e.g. attach new BPMN elements)