5/6 Parallelize OTLP corpus generation; download compressed .pb when available#2138
Open
gareth-ellis wants to merge 5 commits into
Open
5/6 Parallelize OTLP corpus generation; download compressed .pb when available#2138gareth-ellis wants to merge 5 commits into
gareth-ellis wants to merge 5 commits into
Conversation
The ApiError handler in execute_single() decodes `e.body`, `e.error`, and `e.info` as UTF-8 to build a human-readable error message. When the body is binary (e.g., binary protobuf returned by ES OTLP endpoints on 4xx/5xx), the strict decode raises UnicodeDecodeError, which crashes the worker mid-task. Switch the six decode() calls to use errors="replace" so undecodable bytes become U+FFFD instead of aborting the worker. No semantic change for valid UTF-8 (the common case). This is a latent bug independent of OTLP — any operation that surfaces a binary error body would have hit it.
Introduces OtlpProtobufFile in esrally/utils/io.py for reading/writing length-prefixed OTLP ExportMetricsServiceRequest protobufs, plus an offset sidecar to allow worker partitions to seek without scanning. Wires preparation into esrally/track/loader.py and esrally/track/track.py: - New OTLP document set fields (otlp_pb_size_in_bytes, etc.) - prepare_otlp_document_set tries to download a .pb from the corpus base URL, otherwise converts a local JSON corpus to .pb on disk. - set_absolute_data_path picks up the .pb when present. Adds OTLP protobuf bindings to pyproject.toml (opentelemetry-proto). Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
OtlpParamSource streams length-prefixed protobuf records out of an OtlpProtobufFile, partitions them across workers using the offset sidecar, and surfaces percent_completed so the progress bar tracks real progress. Supports a "looped" mode that cycles the partition indefinitely for time-bound benchmarks. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
OtlpIngest POSTs serialized protobuf bytes to the OTLP metrics endpoint, disabling transport-level fast retries in favour of an explicit exponential-with-full-jitter backoff loop. 429/502/503/504 and connection errors are retried; non-retryable ApiErrors return a failure dict so the driver records the error without crashing the worker. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
…lable Adds two perf improvements for OTLP corpora that don't change the on-the-wire benchmark behaviour: 1. Parallel .pb generation. OtlpProtobufFile.create() now uses a ProcessPoolExecutor pipeline (worker count tunable via RALLY_OTLP_CONVERSION_WORKERS) so converting a multi-GB JSON corpus completes in minutes instead of hours. 2. Compressed .pb download. When the JSON corpus is published in a compressed archive, prepare-track first tries the matching compressed .pb (e.g. .pb.zst) from the corpus URL and decompresses locally — typically 2-4x less network bytes than the raw .pb. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
3 tasks
This was referenced May 29, 2026
gareth-ellis
changed the title
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
2 participants
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.
Summary
Two perf improvements for OTLP corpus preparation. Neither changes hot-path benchmark behaviour.
1. Parallel .pb generation.
OtlpProtobufFile.create()now uses aProcessPoolExecutorpipeline — JSON lines stream in, worker processes parse and serialize in parallel, the main process writes batched results in source order. Worker count tunable viaRALLY_OTLP_CONVERSION_WORKERS. Converting an 82 GB JSON corpus drops from hours to minutes.2. Compressed .pb download. When the JSON corpus is published as a compressed archive (e.g.
.zst), prepare-track now tries the matching compressed.pbfrom the corpus base URL (e.g..pb.zst) and decompresses locally. Typically 2–4× less network bytes than the raw.pb.Depends on #2135, #2136, #2137 — merge after all three. Part 5 of 6.
Series
Test plan
🤖 Generated with Claude Code