I am facing an intermittent issue with sub-process execution in Appian.
A sub-process completes successfully and, in the parent process, the sub-process node shows as completed. However, the main process does not advance to the next node and remains stuck at that point.
This is happening both in multiple process models and at different points in the flow.
The same process models have executed successfully in other test runs, including cases where the same sub-process followed the same Flow and completed correctly. There are no errors logged, and the sub-process always completes correctly.
The only common pattern I have noticed is that the issue tends to happen in process instances that have been running for a long time and have traversed many nodes and multiple possible paths before reaching the sub-process.
Has anyone encountered a similar behavior?
Are there known engine limitations or considerations (e.g. long-running process instances, high node count, memory/state handling, or sub-process configuration) that could cause the parent process not to continue after a sub-process completes?
Any guidance on what to check or how to further troubleshoot this would be appreciated.
Discussion posts and replies are publicly visible
This it's the parent process:
And the completed subprocess:
In Appian,very long-running and complex process models can intermittently fail to resume after sub-process completion due to execution state persistence limits. The recommended approach is to minimize process state, reduce node count, and design shorter, event-driven processes.
This feels like you directly copied the business BPMN process into Appian, which is a bad idea. These typically show the whole end-to-end process in a single picture, which leads to long running process instances in Appian and an overall inflexible design.
Try to turn that process into smaller chunks where the first starts the next. Pass only the primary key of the main record to avoid dependencies.
Make sure that user input tasks can be completed within max 3 days.
Process instance state becomes too large, causing silent failure after sub-process completion.Best Practices to Avoid This Issue:Nullify large PVs after use (set to null when no longer needed).Divide long processes into smaller sub-processes.Use async sub-processes where possible.Use database-driven user input tasks instead of process-driven forms (reduces memory and improves performance)Avoid long-running process instances - chain processes instead.I would say proper process design + data management = fewer engine issues.