fix: use task-stream index instead of wall clock in get_task_stream context manager (#9253)#9282
Open
MohammadYusif wants to merge 1 commit into
Open
fix: use task-stream index instead of wall clock in get_task_stream context manager (#9253)#9282MohammadYusif wants to merge 1 commit into
MohammadYusif wants to merge 1 commit into
Conversation
…ontext manager (dask#9253) The get_task_stream context manager bounded the collected tasks with a wall-clock timestamp (time() - 0.1). collect() then bisected the buffer by comparing that boundary against each task's recorded stop time. When there was latency or clock skew between the client and the workers, a task that finished inside the block could carry a stop time earlier than the client's start boundary and be silently dropped, so get_task_stream() returned no tasks. Record the scheduler's monotonic task-stream append index on entry and collect everything appended after it on exit. This removes the dependency on synchronized clocks entirely, as the maintainers' FIXME suggested. Adds a get_task_stream_index scheduler RPC and a start_index path through collect()/get_task_stream(), with tests covering the index semantics and the clock-skew regression.
Contributor
Unit Test ResultsSee test report for an extended history of previous test failures. This is useful for diagnosing flaky tests. 31 files ± 0 31 suites ±0 10h 49m 37s ⏱️ + 14m 21s For more details on these failures, see this check. Results for commit 2d46d3f. ± Comparison against base commit bcad953. This pull request skips 1 test. |
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
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Closes #9253
pre-commit run --all-filesThe
get_task_streamcontext manager bounded the tasks it collected with awall-clock timestamp (
time() - 0.1), andcollect()bisects the buffer bycomparing that boundary against each task's recorded
stoptime. With latencyor clock skew between the client and the workers, a task that finished inside
the block can carry a
stoptime earlier than the client'sstartboundaryand be silently dropped — so
get_task_stream()returns no tasks. This matchesthe existing
# FIXME ... We should query TaskStreamPlugin.index instead.This records the scheduler's monotonic task-stream append index on entry and
collects everything appended after it on exit, removing the dependency on
synchronized clocks.
distributed/diagnostics/task_stream.py:collect()gains astart_indexpath that selects records by append position instead of timestamp.
distributed/scheduler.py: newget_task_stream_indexRPC +start_indexpassthrough on
get_task_stream. Refactored the plugin-init guard into_task_stream_plugin()helper used by both methods.distributed/client.py:get_task_stream/_get_task_streamforwardstart_index; the context manager snapshots the index on enter and collectsfrom it on exit (sync and async), removing the FIXME comment.
distributed/diagnostics/tests/test_task_stream.py: tests for the indexsemantics and the clock-skew regression.