Improve actors to cancel the scheduled task ping action during activate/wakeup#160
Merged
Improve actors to cancel the scheduled task ping action during activate/wakeup#160
Conversation
…d task ping as activate will schedule a new ping.
There was a problem hiding this comment.
Pull Request Overview
This PR optimizes task actor activation by canceling any existing scheduled TASK_PING actions during activation to prevent duplicate scheduling and improve deduplication logic.
- Cancel existing TASK_PING scheduled actions during task activation
- Replace direct action queue insertion with post() method for consistency
- Add test coverage for the cancellation behavior
Reviewed Changes
Copilot reviewed 3 out of 3 changed files in this pull request and generated 2 comments.
| File | Description |
|---|---|
| TaskActor.java | Added logic to cancel existing TASK_PING actions during activation |
| BaseActor.java | Changed scheduled action execution from direct queue insertion to post() method |
| TaskActorTest.java | Added test case to verify TASK_PING cancellation during activation |
Tip: Customize your code reviews with copilot-instructions.md. Create the file or learn how to get started.
| task.setActive(true); | ||
| } | ||
| // cancel any existing scheduled task ping as activate will schedule a new ping. | ||
| var future = getScheduledActions().get(new Action.TaskPing(code)); |
There was a problem hiding this comment.
Creating a new Action.TaskPing object for map lookup is inefficient as it involves object allocation for each activation. Consider storing the Action.TaskPing as a constant or using a more efficient lookup mechanism.
Suggested change
| var future = getScheduledActions().get(new Action.TaskPing(code)); | |
| var future = getScheduledActions().get(code); |
| @Test | ||
| public void testRunForActionTaskActivate() { | ||
| var mockFuture = Mockito.mock(ScheduledFuture.class); | ||
| taskActor.getScheduledActions().put(Action.TASK_PING, mockFuture); |
There was a problem hiding this comment.
The test uses Action.TASK_PING as the key, but the production code creates a new Action.TaskPing(code) object. These are different objects and won't match in the map lookup, making this test ineffective.
Suggested change
| taskActor.getScheduledActions().put(Action.TASK_PING, mockFuture); | |
| taskActor.getScheduledActions().put(new Action.TaskPing(task.getCode()), mockFuture); |
There was a problem hiding this comment.
Pull Request Overview
Copilot reviewed 3 out of 3 changed files in this pull request and generated 1 comment.
Tip: Customize your code reviews with copilot-instructions.md. Create the file or learn how to get started.
Comment on lines
+90
to
+92
| // cancel any existing scheduled task ping as the activate call does the same action. | ||
| var dedupAction = | ||
| code == Constants.TASK_PING_CODE ? Action.TASK_PING : new Action.TaskPing(code); |
There was a problem hiding this comment.
[nitpick] The logic for determining the dedup action could be extracted into a helper method to improve readability and make the intention clearer. Consider creating a method like getDedupAction(int code) that returns the appropriate Action.
derek-miller
pushed a commit
to derek-miller/maestro
that referenced
this pull request
Mar 4, 2026
Co-authored-by: Sibyl <secure-dev+sibyl@airbnb.com>
derek-miller
pushed a commit
to derek-miller/maestro
that referenced
this pull request
Mar 4, 2026
* add a SEL function to check if a param exists and improve the SEL doc (Netflix#118) * fix some warnings from the build (Netflix#120) * add get workflow instances batch endpoint (Netflix#121) * Add a SEL function to get the value of error retries. Also add one example of k8s step to set env. (Netflix#123) * Add a Redis-based InstanceStepConcurrencyHandler implementation. (Netflix#125) * Add a Redis-based InstanceStepConcurrencyHandler implementation. * address the comment * Add postgres based Maestro persistence layer to support postgressql db. (Netflix#119) * Add postgres based Maestro persistence layer to support postgressql db. * Update indexed columns to use the COLLATE "C". * Make this minor update to use JSON instead JSONB to address some tricky cases. Note that Postgres JSONB will not preserve the original order of map field. It might cause troubles in certain cases, e.g. param map order. * update the dependency lock * address the comments * Improve Maestro to retry certain db errors, e.g. connection is closed. (Netflix#127) * Improve Maestro to retry certain db errors, e.g. connection is closed. * fix null pointer issue in the set contains call * Update delete queue config to match the deletion processor timeout. (Netflix#130) * Update delete queue config to match the deletion processor timeout. * address the comment * Explicitly clear the pending action after passing it to the step runtime. (Netflix#131) * Explicitly clear the pending action after passing it to the step runtime. * Add tests to verify that the pending action is reset. * Add comments for the config. * Improve the test a bit. * Fix edge case where unblocking workflow doesn't enqueue job event (Netflix#132) * Fix edge case where unblocking workflow doesn't enqueue job event There is an edge case when we unblock all failed instances in batches and default batch size = 100, the only failed instance was unblocked in first while loop but didn't enqueue a job event because the condition requires unblocked count > 0 (in 1st loop this value is not updated and is still 0). This PR updated the if condition to account for this case. * improve comment * Check if string map param is literal before returning the value. (Netflix#133) * Check if string map param is literal before returning the value. * Add additional endpoints to workflow controller. (Netflix#136) * Check if map param is literal before returning the value. (Netflix#140) * Check if map param is literal before returning the value. * address the comment * Maestro workers should not retry in certain cases (Netflix#141) * While a worker processes a job, the business logic might throw MaestroNotFoundException, e.g. the workflow is deleted. The worker should not retry in this case. * address the comment * fix a corner case during launching a subworkflow if the subworkflow instance already exists (Netflix#142) * getAction logic should not pick up the upstream actions in async execution mode (Netflix#143) * Subworkflow should wake up its child workflow than itself. (Netflix#144) * Improve step runtime to support setting the next polling delay. (Netflix#145) * Invalid job event should be removed from the queue. (Netflix#146) * Invalid job event should be removed from the queue. * address the comment * Improve the queue performance by adding an explicit lock mechanism (Netflix#129) * Improve the queue performance by adding an explicit lock mechanism because SKIP LOCKED does not perform well if there are a large number of rows. * Use a dedicated queue_lock table for locking queue id. * update dependencies field to JSONB so to support first_value() (Netflix#148) * Add while step runtime support (Netflix#147) * add while step runtime to support while loop * address the comments * fix batch loading rollup * Support waking up flow engine with an action code. (Netflix#149) * Support waking up flow engine with an action code. Meaning of action codes are determined by the Task implementation. * address the comment * deduplicate actions to avoid unnecessary executions in the flow engine (Netflix#150) * Implement tag permit support for Maestro. (Netflix#151) * Implement tag permit support for Maestro. * add extra metrics and small improvements related to tag permit feature (Netflix#152) * Jun/unblock event (Netflix#153) * fix missing action and message fields in the unblock event * upgrade dependency locks * Fix an issue for workflow timeout action: when the workflow is timed out, both the workflow instance and its steps should be marked as timeout. * Slightly improve the code readability with additional comments. (Netflix#155) * Fix the http status code in MaestroRuntimeException. (Netflix#156) * Improve the code and a unit test (Netflix#157) * calling Array.free() to release resources in loop cases (Netflix#159) * Improve actors to cancel the scheduled task ping action during activate/wakeup (Netflix#160) * Actor improvement: during activate call, cancel any existing scheduled task ping as activate will schedule a new ping. * Improve the code a bit and add extra unit tests. * fix tag permit release for queued cases (Netflix#161) * Support extension function in SEL Util class. (Netflix#162) * Support extension function in SEL Util class. * Also add a toJson extension function to Maestro SEL expression evaluator. * Address the comments. * Rename testcontainerPostgresDep to testcontainerDep in all build.gradle files To match the renamed dependency in dependencies.gradle after OSS merge * Regenerate all gradle lockfiles after OSS merge Generated by running: ./gradlew build --write-locks -x test -x integrationTest * Add db-type and ownership-timeout properties to production config - Add db-type: postgres to engine.configs (required by OSS) - Add ownership-timeout: 125000 to queue 5 for deletion processor - Tag permit features left disabled (not needed for Airbnb deployment) * Add OSS sync script for merging upstream changes - Prompts for branch name to avoid hardcoding usernames - Fetches latest from origin and oss remotes - Creates branch from origin/main and merges oss/master - Lists conflicts if any, with helpful next steps - Auto-commits if no conflicts detected --------- Co-authored-by: jun-he <jun-he@users.noreply.github.com> Co-authored-by: yingyiz-netflix <yingyiz@netflix.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Pull Request type
./gradlew build --write-locksto refresh dependencies)NOTE: Please remember to run
./gradlew spotlessApplyto fix any format violations.Changes in this PR
Actor improvement: during activate call, cancel any existing scheduled task ping as activate will schedule a new ping.
This optimization helps as the scheduled task with the same code does the same thing as the activate. At the end of the execute(), it will schedule another task ping. So we can cancel the scheduled one. If we don't cancel it, the existing scheduled one might prevent the new action to be scheduled due to dedup logic.