FEO Error Handling #944

johannes-esr · 2025-04-15T13:48:17Z

johannes-esr
Apr 15, 2025
Collaborator

Participants

@johannes-esr
@armin-acn
@MarkSchmitt
@vinodreddy-g
@pawelrutkaq
@ramceb

Agenda

https://github.com/eclipse-score/inc_feo/issues/23
- Errors arising in FEO Applications
- Error handling on platform level e.g. Health Management.

Meeting Minutes

What happens when one FEO activity/process fails?
What happens to the other processes?
What requirements regarding (restart-)time are needed?
How to safely shut down a taskchain?
What happens if activities call middleware and an error occurs in middleware (error passed back to activity OR middleware panics)?
Error Cases:
- Independent from state:
  - Possible fault: primary process dies
  - Possible fault: secondary process dies
- Lifecycle Manger creates all processes (primary & secondaries)
  - Possible fault: processes cannot be created (Possible resolution: Lifecycle Manager have to deal with it (e.g. restart / try again))
- Lifecycle Manager creates all processes (primary & secondaries)
  - Possible fault: not all secondaries report back to the primary
  - Possible fault: timeout during connection of secondaries
- Lifecycle Manager creates all processes (primary & secondaries), all secondaries report back to the primary, primary calls startup() of all activities
  - Possible fault: one of the activities fails in the startup() function, not all activities have been properly initialized
  - Possible fault: timeout during startup of activities
- Lifecycle Manager creates all processes (primary & secondaries), all secondaries report back to the primary, primary calls startup() of all activities, task chain is run
  - Possible fault: one of the activities fails in the step() function
  - Possible fault: timeout (one activity takes too long in the step() function OR secondary doesn't signal back to primary)
  - Possible fault: cycle time not met (whole task chain takes too long)
  - Possible fault: intermediate (time/memory/cpu-)budgets for activities are not met (to be defined whether this is a desired feature, "supervision")
- Lifecycle Manager creates all processes (primary & secondaries), all secondaries report back to the primary, primary calls startup() of all activities, task chain is stopped
  - Possible fault: one of the activities fails in the shutdown() function, not all activities have been properly shut down

ramceb · 2025-05-22T14:06:30Z

ramceb
May 22, 2025
Collaborator

Error Cases:
- Independent from state:
  - REQ 1: If the primary process dies the external lifecycle management shall kill all dependent processes.
  - REQ 2: If the secondary process dies the lifecycle management shall send a SIGTERM to the primary process. The primary process shall send a shutdown signal to all remaining activities.
- Lifecycle Manger creates all processes (primary & secondaries)
  - Possible fault: processes cannot be created (Possible resolution: Lifecycle Manager have to deal with it (e.g. restart / try again))
- Lifecycle Manager creates all processes (primary & secondaries)
  - REQ 3: If not all secondaries report back to the primary in time, the primary shall report an error to the lifecycle/health management. The startup functions shall not be triggered.
- Lifecycle Manager creates all processes (primary & secondaries), all secondaries report back to the primary, primary calls startup() of all activities
  - REQ 4: If an error occurs during execution of a startup function, the primary process shall abort calling startup functions, trigger healthmanagement and terminate itself. For all activities where the startup has already been triggered successfully the corresponding shutdown function shall be executed in arbitrary sequence.
  - TODO: check if return value of startup/step/shutdown function is needed
  - Possible fault: timeout during startup of activities
  - REQ 5: Any timeout during startup, run or shutdown of activities shall be reported to the health-management. The primary process shall initiate a shutdown for all successfully started activities and terminate itself.
- Lifecycle Manager creates all processes (primary & secondaries), all secondaries report back to the primary, primary calls startup() of all activities, task chain is run
  REQ 6: If one activity fails in the step function, a logical waypoint error shall be reported to health management. The primary process shall initiate a shutdown for all successfully started activities and terminate itself.
  - Possible fault: timeout (one activity takes too long in the step() function OR secondary doesn't signal back to primary) ==> see REQ 5
  - Possible fault: intermediate (time/memory/cpu-)budgets for activities are not met (to be defined whether this is a desired feature, "supervision"). ==> Resource supervision and quotas shall be handled in s separate feature request.
- Lifecycle Manager creates all processes (primary & secondaries), all secondaries report back to the primary, primary calls startup() of all activities, task chain is stopped
  - Possible fault: one of the activities fails in the shutdown() function, not all activities have been properly shut down
    REQ 7: If one activity fails in the shutdown function, a logical waypoint error shall be reported to health management. The primary process shall shutdown all remaining activities and terminate itself.

0 replies

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Eclipse S-CORE

FEO Error Handling #944

Uh oh!

{{title}}

Uh oh!

Replies: 1 comment

Uh oh!

{{title}}

Uh oh!

Select a reply

Uh oh!

Eclipse S-CORE

FEO Error Handling #944

Uh oh!

johannes-esr Apr 15, 2025 Collaborator

Participants

Agenda

Meeting Minutes

Replies: 1 comment

Uh oh!

ramceb May 22, 2025 Collaborator

johannes-esr
Apr 15, 2025
Collaborator

ramceb
May 22, 2025
Collaborator