feat: support terminate end events #10442

saig0 · 2022-09-22T08:19:02Z

Description

Add support for BPMN terminate end events. See #8789 (comment) on how the BPMN element should work.

The implementation doesn't follow the BPMN spec in one point: the flow scope that contains the terminate end event is not terminated but completed. Reasoning:

The state of the flow scope is a detail that doesn't influence the core behavior. In both cases, the process instance should continue, for example, by taking the outgoing sequence flow. The difference is not visible to process participants but only when monitoring the process instance, for example, in Operate.
It fits better with the existing implementation. It would be a bigger effort to continue the process instance (e.g. taking the outgoing sequence flow) when the flow scope is terminated. As a result, we would end up in more complex code.
It aligns with the behavior of Camunda Platform 7.

Side note: I implemented the major parts during a Live Hacking session. 🎥

Related issues

closes #8789

Definition of Done

Not all items need to be done depending on the issue and the pull request.

Code changes:

The changes are backwards compatibility with previous versions
If it fixes a bug then PRs are created to backport the fix to the last two minor versions. You can trigger a backport by assigning labels (e.g. backport stable/1.3) to the PR, in case that fails you need to create backports manually.

Testing:

There are unit/integration tests that verify all acceptance criterias of the issue
New tests are written to ensure backwards compatibility with further versions
The behavior is tested manually
The change has been verified by a QA run
The impact of the changes is verified by a benchmark

Documentation:

The documentation is updated (e.g. BPMN reference, configuration, examples, get-started guides, etc.)
New content is added to the release announcement
If the PR changes how BPMN processes are validated (e.g. support new BPMN element) then the Camunda modeling team should be informed to adjust the BPMN linting.

Please refer to our review guidelines.

github-actions · 2022-09-22T08:49:47Z

Test Results

  927 files +  2   927 suites +2 2h 5m 26s ⏱️ + 5m 17s
7 448 tests +73 7 438 ✔️ +73 10 💤 ±0 0 ❌ ±0
7 636 runs +73 7 626 ✔️ +73 10 💤 ±0 0 ❌ ±0

Results for commit 1668ad1. ± Comparison against base commit c7baaf4.

♻️ This comment has been updated with latest results.

saig0 · 2022-09-22T10:39:24Z

Note that the failing test case is not related to the changes. I can reproduce the failure also on main. I created an issue for further analysis: #10445.

saig0 · 2022-09-22T11:14:12Z

✔️ I tested the terminate end event manually in Operate.

remcowesterhoud

Whew that was big one 😄 Nice work @saig0 🚀

I have a few questions and comments. But my main thing is what you mentioned in the PR.

The implementation doesn't follow the BPMN spec in one point: the flow scope that contains the terminate end event is not terminated but completed.

❌ I think we should reconsider this. In my opinion the downsides don't outweigh the positives by terminating instead of completing.

You mention that this decision doesn't influence the core behavior. While this is true, it does cause a strange behavior for users willing to test their process using zeebe-process-test. For example, we have an assertion verifying if a process is terminated. For users it would be natural to verify this in the case of a terminate end event. However, this assertion will fail, because in reality Zeebe will complete the process.
You mention it fits better with the existing implementation. I have no doubts that this is true and you will definitely know better than me 😄. The code will become more complex, but the behavior of the code would do what feels natural and is expected. I noticed I had a bit of trouble reviewing this PR already, because sometimes an element would be terminated, other times it would be completed. I understand how it works after the review, but it feels unpredictable, which is again something that consumers of the exported records may have trouble with.
You mention it aligns better with C7. So let's improve it 😄

Please reconsider this approach.

❌ It would be cool if we can support this terminate end event in our randomised tests. This can be a separate issue which we can work on later.

...ne/src/test/java/io/camunda/zeebe/engine/processing/bpmn/activity/TerminateEndEventTest.java

engine/src/main/java/io/camunda/zeebe/engine/processing/bpmn/event/EndEventProcessor.java

engine/src/main/java/io/camunda/zeebe/engine/processing/bpmn/container/ProcessProcessor.java

remcowesterhoud · 2022-09-23T11:10:28Z

...ne/src/test/java/io/camunda/zeebe/engine/processing/bpmn/activity/TerminateEndEventTest.java

+            .multiInstance(
+                multiInstance ->
+                    multiInstance
+                        .parallel()


🔧 I think we should also test the sequential multi instance.

remcowesterhoud

🍪

saig0 · 2022-09-23T13:03:57Z

bors merge

saig0 · 2022-09-23T13:11:49Z

We discussed if the flow scope should be completed or terminated. Besides the spec, we have different options about the user's (or our) expectations. From the code side, we agreed that the completion behavior fits better and avoids more complex logic for the particular case of the terminate end event.

We decided to complete the flow scope for now. We could change the behavior later if we see a strong need for it. Or, if it aligns with further implementation of other BPMN elements. From a user point-of-view, it's mainly a change of the icon in Operate.

10442: feat: support terminate end events r=saig0 a=saig0 ## Description Add support for BPMN terminate end events. See #8789 (comment) on how the BPMN element should work. The implementation doesn't follow the BPMN spec in one point: the flow scope that contains the terminate end event is not terminated but completed. Reasoning: - The state of the flow scope is a detail that doesn't influence the core behavior. In both cases, the process instance should continue, for example, by taking the outgoing sequence flow. The difference is not visible to process participants but only when monitoring the process instance, for example, in Operate. - It fits better with the existing implementation. It would be a bigger effort to continue the process instance (e.g. taking the outgoing sequence flow) when the flow scope is terminated. As a result, we would end up in more complex code. - It aligns with the behavior of Camunda Platform 7. Side note: I implemented the major parts during a [Live Hacking session](https://www.twitch.tv/videos/1584245006). 🎥 ## Related issues closes #8789 Co-authored-by: Philipp Ossler <philipp.ossler@gmail.com>

zeebe-bors-camunda · 2022-09-23T15:38:35Z

Build failed:

Test summary

saig0 · 2022-09-23T17:48:04Z

Bors merge

10442: feat: support terminate end events r=saig0 a=saig0 ## Description Add support for BPMN terminate end events. See #8789 (comment) on how the BPMN element should work. The implementation doesn't follow the BPMN spec in one point: the flow scope that contains the terminate end event is not terminated but completed. Reasoning: - The state of the flow scope is a detail that doesn't influence the core behavior. In both cases, the process instance should continue, for example, by taking the outgoing sequence flow. The difference is not visible to process participants but only when monitoring the process instance, for example, in Operate. - It fits better with the existing implementation. It would be a bigger effort to continue the process instance (e.g. taking the outgoing sequence flow) when the flow scope is terminated. As a result, we would end up in more complex code. - It aligns with the behavior of Camunda Platform 7. Side note: I implemented the major parts during a [Live Hacking session](https://www.twitch.tv/videos/1584245006). 🎥 ## Related issues closes #8789 Co-authored-by: Philipp Ossler <philipp.ossler@gmail.com>

zeebe-bors-camunda · 2022-09-23T18:08:49Z

Build failed:

Test summary

saig0 · 2022-09-26T03:29:04Z

bors merge

10442: feat: support terminate end events r=saig0 a=saig0 ## Description Add support for BPMN terminate end events. See #8789 (comment) on how the BPMN element should work. The implementation doesn't follow the BPMN spec in one point: the flow scope that contains the terminate end event is not terminated but completed. Reasoning: - The state of the flow scope is a detail that doesn't influence the core behavior. In both cases, the process instance should continue, for example, by taking the outgoing sequence flow. The difference is not visible to process participants but only when monitoring the process instance, for example, in Operate. - It fits better with the existing implementation. It would be a bigger effort to continue the process instance (e.g. taking the outgoing sequence flow) when the flow scope is terminated. As a result, we would end up in more complex code. - It aligns with the behavior of Camunda Platform 7. Side note: I implemented the major parts during a [Live Hacking session](https://www.twitch.tv/videos/1584245006). 🎥 ## Related issues closes #8789 Co-authored-by: Philipp Ossler <philipp.ossler@gmail.com>

zeebe-bors-camunda · 2022-09-26T03:52:37Z

Build failed:

Test summary

saig0 · 2022-09-26T03:57:29Z

bors merge

10442: feat: support terminate end events r=saig0 a=saig0 ## Description Add support for BPMN terminate end events. See #8789 (comment) on how the BPMN element should work. The implementation doesn't follow the BPMN spec in one point: the flow scope that contains the terminate end event is not terminated but completed. Reasoning: - The state of the flow scope is a detail that doesn't influence the core behavior. In both cases, the process instance should continue, for example, by taking the outgoing sequence flow. The difference is not visible to process participants but only when monitoring the process instance, for example, in Operate. - It fits better with the existing implementation. It would be a bigger effort to continue the process instance (e.g. taking the outgoing sequence flow) when the flow scope is terminated. As a result, we would end up in more complex code. - It aligns with the behavior of Camunda Platform 7. Side note: I implemented the major parts during a [Live Hacking session](https://www.twitch.tv/videos/1584245006). 🎥 ## Related issues closes #8789 Co-authored-by: Philipp Ossler <philipp.ossler@gmail.com>

zeebe-bors-camunda · 2022-09-26T04:21:52Z

Build failed:

Test summary

saig0 · 2022-09-26T04:24:47Z

bors merge

10442: feat: support terminate end events r=saig0 a=saig0 ## Description Add support for BPMN terminate end events. See #8789 (comment) on how the BPMN element should work. The implementation doesn't follow the BPMN spec in one point: the flow scope that contains the terminate end event is not terminated but completed. Reasoning: - The state of the flow scope is a detail that doesn't influence the core behavior. In both cases, the process instance should continue, for example, by taking the outgoing sequence flow. The difference is not visible to process participants but only when monitoring the process instance, for example, in Operate. - It fits better with the existing implementation. It would be a bigger effort to continue the process instance (e.g. taking the outgoing sequence flow) when the flow scope is terminated. As a result, we would end up in more complex code. - It aligns with the behavior of Camunda Platform 7. Side note: I implemented the major parts during a [Live Hacking session](https://www.twitch.tv/videos/1584245006). 🎥 ## Related issues closes #8789 Co-authored-by: Philipp Ossler <philipp.ossler@gmail.com>

zeebe-bors-camunda · 2022-09-26T04:51:16Z

Build failed:

Test summary

saig0 · 2022-09-26T04:52:42Z

bors merge

10442: feat: support terminate end events r=saig0 a=saig0 ## Description Add support for BPMN terminate end events. See #8789 (comment) on how the BPMN element should work. The implementation doesn't follow the BPMN spec in one point: the flow scope that contains the terminate end event is not terminated but completed. Reasoning: - The state of the flow scope is a detail that doesn't influence the core behavior. In both cases, the process instance should continue, for example, by taking the outgoing sequence flow. The difference is not visible to process participants but only when monitoring the process instance, for example, in Operate. - It fits better with the existing implementation. It would be a bigger effort to continue the process instance (e.g. taking the outgoing sequence flow) when the flow scope is terminated. As a result, we would end up in more complex code. - It aligns with the behavior of Camunda Platform 7. Side note: I implemented the major parts during a [Live Hacking session](https://www.twitch.tv/videos/1584245006). 🎥 ## Related issues closes #8789 Co-authored-by: Philipp Ossler <philipp.ossler@gmail.com>

zeebe-bors-camunda · 2022-09-26T05:13:18Z

Build failed:

Test summary

saig0 · 2022-09-26T05:24:06Z

bors merge

10442: feat: support terminate end events r=saig0 a=saig0 ## Description Add support for BPMN terminate end events. See #8789 (comment) on how the BPMN element should work. The implementation doesn't follow the BPMN spec in one point: the flow scope that contains the terminate end event is not terminated but completed. Reasoning: - The state of the flow scope is a detail that doesn't influence the core behavior. In both cases, the process instance should continue, for example, by taking the outgoing sequence flow. The difference is not visible to process participants but only when monitoring the process instance, for example, in Operate. - It fits better with the existing implementation. It would be a bigger effort to continue the process instance (e.g. taking the outgoing sequence flow) when the flow scope is terminated. As a result, we would end up in more complex code. - It aligns with the behavior of Camunda Platform 7. Side note: I implemented the major parts during a [Live Hacking session](https://www.twitch.tv/videos/1584245006). 🎥 ## Related issues closes #8789 Co-authored-by: Philipp Ossler <philipp.ossler@gmail.com>

zeebe-bors-camunda · 2022-09-26T05:45:28Z

Build failed:

Test summary

Extend the end event builder to add a terminate event definition. The event definition morphs the event into a terminate end event.

If the process instance reach a terminate end event on the root scope then we should complete the process instance.

Extract the end event validation tests into a separate class. Morph the test cases in JUnit 5.

Morph the end event tests into parameterized tests for all supported and unsupported types of end events.

If the process instance reaches a terminate end event in the root scope then we should terminate all active element instances. After the element instances are terminated, we should complete the process instance.

Extend the executable model for end events to identify terminate end events. If the terminate end event is completed then we terminate all other element instances of the same flow scope.

The process instance reached a terminate end event in the process that terminated all element instances of the process. After all element instances are terminated, we complete the process instance. We check the current state of the process instance to distingulish the behavior of a regular cancallation of the process instance, or an interruption from a parent process instance.

If the process instance reaches a terminate end event in a subprocess then we should terminate all active element instances inside the subprocess. After the element instances are terminated, we should complete the subprocess and take the outgoing sequence flow. Other element instances outside the subprocess should not be terminated.

The process instance reached a terminate end event in the subprocess that terminated all element instances inside the subprocess. After all element instances are terminated, we complete the subprocess.

If the process instance reaches a terminate end event in an event subprocess then we should terminate all active element instances inside the event subprocess. After the element instances are terminated, we should complete the event subprocess. Other element instances outside the event subprocess should not be terminated.

The process instance reached a terminate end event in the event subprocess that terminated all element instances inside the event subprocess. After all element instances are terminated, we complete the event subprocess.

If the process instance reaches a terminate end event in a child process then we should terminate all active element instances of the child process. After the element instances are terminated, we should complete the child process instance and complete the call activity.

If the process instance reaches a terminate end event in a multi-instance subprocess then we should terminate all active element instances inside this subprocess. After the element instances are terminated, we should complete the subprocess. Other instances of the multi-instance should not be terminated.

saig0 · 2022-09-26T08:08:36Z

bors merge

10442: feat: support terminate end events r=saig0 a=saig0 ## Description Add support for BPMN terminate end events. See #8789 (comment) on how the BPMN element should work. The implementation doesn't follow the BPMN spec in one point: the flow scope that contains the terminate end event is not terminated but completed. Reasoning: - The state of the flow scope is a detail that doesn't influence the core behavior. In both cases, the process instance should continue, for example, by taking the outgoing sequence flow. The difference is not visible to process participants but only when monitoring the process instance, for example, in Operate. - It fits better with the existing implementation. It would be a bigger effort to continue the process instance (e.g. taking the outgoing sequence flow) when the flow scope is terminated. As a result, we would end up in more complex code. - It aligns with the behavior of Camunda Platform 7. Side note: I implemented the major parts during a [Live Hacking session](https://www.twitch.tv/videos/1584245006). 🎥 ## Related issues closes #8789 10443: Do not take a backup if it already exists r=deepthidevaki a=deepthidevaki ## Description After restore, the log is truncated to the checkpoint position. So the checkpoint record is processed again and will trigger a new backup with the same Id of the backup it restored from. With this PR, `BackupService` handles this case gracefully. In addition, we also do not take a new backup if existing backup is failed or in progress. Alternatively, we can delete this backup and take a new one. But chances of it happening (i.e triggering a new backup when one already is in progress/failed) is very low. So we can keep this simple. ## Related issues closes #10430 Co-authored-by: Philipp Ossler <philipp.ossler@gmail.com> Co-authored-by: Deepthi Devaki Akkoorath <deepthidevaki@gmail.com>

zeebe-bors-camunda · 2022-09-26T08:55:56Z

Build failed (retrying...):

Test summary

zeebe-bors-camunda · 2022-09-26T10:18:46Z

Build succeeded:

saig0 force-pushed the 8789-terminate-end-event branch from 9cf4daa to f7aa26c Compare September 22, 2022 08:21

saig0 marked this pull request as ready for review September 22, 2022 10:39

saig0 requested a review from remcowesterhoud September 22, 2022 10:39

saig0 mentioned this pull request Sep 22, 2022

I can model terminate end events for Camunda Platform 8 camunda/camunda-modeler#3149

Closed

2 tasks

remcowesterhoud suggested changes Sep 23, 2022

View reviewed changes

remcowesterhoud approved these changes Sep 23, 2022

View reviewed changes

saig0 added 18 commits September 26, 2022 10:08

feat(bpmn-model): extend the end event builder

2645df7

Extend the end event builder to add a terminate event definition. The event definition morphs the event into a terminate end event.

test(engine): verify to complete an instance with terminate end event

a02ddcc

If the process instance reach a terminate end event on the root scope then we should complete the process instance.

feat(bpmn-model): add support for terminate end events

c734ae9

refactoring(bpmn-model): extract end event validation tests

f6ad2a4

Extract the end event validation tests into a separate class. Morph the test cases in JUnit 5.

test(bpmn-model): verify that terminate end events are supported

140242e

refactor(bpmn-model): morph into parameterized test for end events

96a6f1a

Morph the end event tests into parameterized tests for all supported and unsupported types of end events.

test(engine): verify the cancellation behavior of terminate end events

bfc2399

If the process instance reaches a terminate end event in the root scope then we should terminate all active element instances. After the element instances are terminated, we should complete the process instance.

feat(engine): add a termination behavior for end events

87af1cc

Extend the executable model for end events to identify terminate end events. If the terminate end event is completed then we terminate all other element instances of the same flow scope.

refactor(engine): bundle the predicate in the end event behavior

ff6bf71

refactor(engine): add static import for tuple

a2e98d8

feat(engine): the subprocess handles terminate end events

3bf0ae5

The process instance reached a terminate end event in the subprocess that terminated all element instances inside the subprocess. After all element instances are terminated, we complete the subprocess.

feat(engine): the event subprocess handles terminate end events

eccfbd9

The process instance reached a terminate end event in the event subprocess that terminated all element instances inside the event subprocess. After all element instances are terminated, we complete the event subprocess.

refactor(engine): extract method to get the kind of end event

1668ad1

saig0 force-pushed the 8789-terminate-end-event branch from 186faea to 1668ad1 Compare September 26, 2022 08:08

zeebe-bors-camunda bot merged commit 721b6b4 into main Sep 26, 2022

zeebe-bors-camunda bot deleted the 8789-terminate-end-event branch September 26, 2022 10:18

lzgabel mentioned this pull request Nov 8, 2022

I can model link events for Camunda Platform 8 camunda/camunda-modeler#3282

Closed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

feat: support terminate end events #10442

feat: support terminate end events #10442

saig0 commented Sep 22, 2022 •

edited

github-actions bot commented Sep 22, 2022 •

edited

saig0 commented Sep 22, 2022

saig0 commented Sep 22, 2022

remcowesterhoud left a comment

remcowesterhoud Sep 23, 2022

remcowesterhoud left a comment

saig0 commented Sep 23, 2022

saig0 commented Sep 23, 2022

zeebe-bors-camunda bot commented Sep 23, 2022

saig0 commented Sep 23, 2022

zeebe-bors-camunda bot commented Sep 23, 2022

saig0 commented Sep 26, 2022

zeebe-bors-camunda bot commented Sep 26, 2022

saig0 commented Sep 26, 2022

zeebe-bors-camunda bot commented Sep 26, 2022

saig0 commented Sep 26, 2022

zeebe-bors-camunda bot commented Sep 26, 2022

saig0 commented Sep 26, 2022

zeebe-bors-camunda bot commented Sep 26, 2022

saig0 commented Sep 26, 2022

zeebe-bors-camunda bot commented Sep 26, 2022

saig0 commented Sep 26, 2022

zeebe-bors-camunda bot commented Sep 26, 2022

zeebe-bors-camunda bot commented Sep 26, 2022

feat: support terminate end events #10442

feat: support terminate end events #10442

Conversation

saig0 commented Sep 22, 2022 • edited

Description

Related issues

Definition of Done

github-actions bot commented Sep 22, 2022 • edited

Test Results

saig0 commented Sep 22, 2022

saig0 commented Sep 22, 2022

remcowesterhoud left a comment

Choose a reason for hiding this comment

remcowesterhoud Sep 23, 2022

Choose a reason for hiding this comment

remcowesterhoud left a comment

Choose a reason for hiding this comment

saig0 commented Sep 23, 2022

saig0 commented Sep 23, 2022

zeebe-bors-camunda bot commented Sep 23, 2022

saig0 commented Sep 23, 2022

zeebe-bors-camunda bot commented Sep 23, 2022

saig0 commented Sep 26, 2022

zeebe-bors-camunda bot commented Sep 26, 2022

saig0 commented Sep 26, 2022

zeebe-bors-camunda bot commented Sep 26, 2022

saig0 commented Sep 26, 2022

zeebe-bors-camunda bot commented Sep 26, 2022

saig0 commented Sep 26, 2022

zeebe-bors-camunda bot commented Sep 26, 2022

saig0 commented Sep 26, 2022

zeebe-bors-camunda bot commented Sep 26, 2022

saig0 commented Sep 26, 2022

zeebe-bors-camunda bot commented Sep 26, 2022

zeebe-bors-camunda bot commented Sep 26, 2022

saig0 commented Sep 22, 2022 •

edited

github-actions bot commented Sep 22, 2022 •

edited