feat: ES throughput + swimlane dashboard with testing infrastructure#912
Merged
Conversation
- Add workers, batch_target_bytes, batch_timeout_ms to pipeline config - Pass PipelineMetrics into OutputWorkerPool via WorkerConfig so workers call finish_active_batch directly after process_item — avoids stale active_batches when the ack select loop is starved by flush_batch - Diagnostics: expose active_batches and per-batch trace data via API - Fix criterion::black_box deprecation in benches (std::hint::black_box) Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
…ck, bar fix - Swimlane: extract layoutSwimlane() pure function for testability - Swimlane: clamp scan barEndMs to nowMs — fixes bar glued to right edge when server/client clock skew or stale xfmMs pushes barEndMs past now - Chart: add [series] dep to plot-creation useEffect — fixes chart destroying and recreating on every parent re-render - Window picker: add 5s option; auto-select based on avg batch total_ns; lock selection once user clicks (userPickedWindowRef) - Add PipelineView, LogViewer, MetricBadges, StatusBar, ConfigView components - Expose /api/pipelines, /api/traces, /api/logs, /api/config endpoints Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- Vitest 4 with happy-dom, v8 coverage, fast-check property tests - Biome v2 formatter + linter; oxlint 1.0 primary linter; tsgo type check - 168 tests across format, ring, rates, stats, layout, lanes modules - CI: frontend-checks (biome/oxlint/tsgo in parallel) + frontend-test jobs - Fix all pre-existing lint errors: useButtonType, noSvgWithoutTitle, useIterableCallbackReturn, useExhaustiveDependencies Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
|
Warning Rate limit exceeded
Your organization is not enrolled in usage-based pricing. Contact your admin to enable usage-based pricing to continue reviews beyond the rate limit, or try again in 5 minutes and 57 seconds. ⌛ How to resolve this issue?After the wait time has elapsed, a review can be triggered using the We recommend that you space out your commits to avoid hitting the rate limit. 🚦 How do rate limits work?CodeRabbit enforces hourly rate limits for each developer per organization. Our paid plans have higher rate limits than the trial, open-source and free plans. In all cases, we re-allow further reviews after a brief timeout. Please see our FAQ for further information. ℹ️ Review info⚙️ Run configurationConfiguration used: Repository YAML (base), Organization UI (inherited) Review profile: ASSERTIVE Plan: Pro Run ID: 📒 Files selected for processing (9)
WalkthroughThis PR implements end-to-end batch instrumentation across the pipeline, worker pool, and diagnostics layers. It adds configurable worker pool sizing and batch parameters to Possibly related PRs
Caution Pre-merge checks failedPlease resolve all errors before merging. Addressing warnings is optional.
❌ Failed checks (1 error, 2 warnings)
✅ Passed checks (2 passed)
Comment |
Owner
Author
|
@coderabbitai review |
✅ Actions performedReview triggered.
|
There was a problem hiding this comment.
Actionable comments posted: 16
Caution
Some comments are outside the diff and can’t be posted inline due to platform limitations.
⚠️ Outside diff range comments (1)
crates/logfwd/src/worker_pool.rs (1)
476-515:⚠️ Potential issue | 🟠 MajorExport the worker-pool queue wait instead of dropping it.
queue_wait_nsis measured here, but it never gets attached to a span or serialized. Thequeue_wait_nsthat currently reaches/api/tracesstill comes fromcrates/logfwd/src/pipeline.rsLines 573-575 /crates/logfwd-io/src/diagnostics.rsLines 850-852, which is input-channel wait before scan, not worker-pool wait between transform and output. Sincetotal_nsincludes the latter, the dashboard's queue breakdown is wrong whenever a batch sits behind another worker job.🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In `@crates/logfwd/src/worker_pool.rs` around lines 476 - 515, queue_wait_ns is measured in the worker loop but never propagated; add it to the acknowledgement path so downstream pipeline/diagnostics see the worker-pool wait. Update the AckItem type to include queue_wait_ns, populate that field in the worker code where ack_tx.send(AckItem { ... }) is called (the block that calls process_item and metrics.finish_active_batch), and ensure the pipeline/diagnostics consumers (where AckItem is handled) read and serialize/attach queue_wait_ns to spans/metrics (e.g. the code paths in pipeline.rs and diagnostics.rs that currently record input-channel wait). Also consider setting the output span field (send_ns/recv_ns or a new queue_wait field) via tracing when creating output_span so traces reflect the worker-pool queue wait.
🤖 Prompt for all review comments with AI agents
Verify each finding against the current code and only fix it if needed.
Inline comments:
In @.github/workflows/ci.yml:
- Around line 77-80: The workflow uses npx tsgo --noEmit directly which relies
on a transitive lockfile entry; add tsgo to dashboard package.json
devDependencies (e.g., "tsgo": "<pinned-version>") so the tool is an explicit
devDependency, or alternatively modify the CI step to run the package script
(e.g., npm run typecheck) that invokes tsgo from package.json; update either the
devDependencies entry for tsgo or the workflow run line to npm run typecheck and
ensure the package.json contains the corresponding "typecheck" script.
In `@bench/scenarios/self-bench.yaml`:
- Line 4: The SQL in the benchmark scenario uses a non-standard column "level"
in the WHERE clause which can break when input schemas lack that field; update
the query in bench/scenarios/self-bench.yaml to use the benchmark-standard
column "level_str" (e.g., replace "level != 'DEBUG'" with "level_str !=
'DEBUG'") so it matches other benchmark/transform SQL and avoids missing-column
errors.
In `@crates/logfwd-config/src/lib.rs`:
- Around line 291-296: The config currently allows workers: 0 which later
triggers assert!(max_workers >= 1) in OutputWorkerPool::new; update config
validation when loading/parsing the struct that contains the workers field to
reject zero values (e.g., return a validation error or map 0 to a meaningful
default) so users get a clear config error instead of a panic. Specifically, add
a check for the workers field (pub workers: Option<usize>) in the
config-load/validate path and ensure Any Some(0) is treated as invalid (with an
Err containing a clear message) or normalized to a minimum of 1 before passing
to OutputWorkerPool::new. Ensure error text references "workers" so callers can
fix the config.
In `@crates/logfwd-io/src/diagnostics.rs`:
- Around line 119-132: The ActiveBatch struct uses sentinel values (worker_id =
-1, output_start_unix_ns = 0) and a stringly-typed stage; change worker_id to
Option<i64> and output_start_unix_ns to Option<u64> so absence is explicit, and
replace stage: &'static str with a Stage enum (e.g. Scan/Transform/Output) to
avoid string comparisons; update ActiveBatch, its
constructor/serializers/deserializers, and any code that reads these fields
(consumers/tests) to handle Option and match on Stage instead of using magic
numeric/special-case checks.
In `@crates/logfwd-output/src/elasticsearch.rs`:
- Around line 136-146: The extract_took code is too strict: after finding PREFIX
("took":) it assumes digits follow immediately, so formatted JSON like `"took":
5` returns None; update extract_took to skip any ASCII whitespace after PREFIX
before scanning for digits (i.e., advance the slice referenced by rest past
is_ascii_whitespace bytes), then find the run of digits and parse that substring
as before; keep using the existing PREFIX, rest, and end variables but ensure
rest begins at the first non-whitespace byte.
In `@crates/logfwd/src/pipeline.rs`:
- Around line 251-257: Validate the YAML knobs before constructing the pool:
ensure that when computing (max_workers, idle_timeout) you check config.workers
is > 0 (unless factory.is_single_use() forces 1) and return a config error
instead of passing 0 into OutputWorkerPool::new; likewise validate
config.batch_target_bytes > 0 before starting input_poll_loop and error out
early (don't let input_poll_loop treat len() >= 0 as always-sendable). Locate
the logic around factory.is_single_use(), OutputWorkerPool::new, config.workers,
and the input_poll_loop and add early validation/Err return (or propagate a
ConfigError) for invalid values so the function fails fast instead of panicking
or spinning.
In `@dashboard/package.json`:
- Line 33: The dependency "@typescript/native-preview" is set to "latest", which
makes installs non-deterministic; update the dashboard package.json to pin this
dependency to a specific nightly version or semver range (for example a nightly
tag like "^7.0.0-dev.20260403") or otherwise ensure reproducible installs by
committing the lockfile and using npm ci in CI; modify the
"@typescript/native-preview" entry accordingly so the dependency version is
deterministic.
In `@dashboard/src/app.tsx`:
- Around line 396-407: The inline filter calls passed to ChartGrid create new
arrays every render; wrap each filtered series in useMemo to return stable
references (e.g., const pipelineSeries = useMemo(() =>
seriesRef.current.filter(s =>
["lps","bps","obps","err","lat","inflight","batches","stalls"].includes(s.id)),
[seriesRef.current]) and const systemSeries = useMemo(() =>
seriesRef.current.filter(s => ["cpu","mem"].includes(s.id)),
[seriesRef.current]) and then pass pipelineSeries and systemSeries to the
respective ChartGrid components so effects/memos in ChartGrid see stable series
identity.
In `@dashboard/src/components/Chart.tsx`:
- Around line 61-69: The effect that creates/destroys the plot currently only
depends on the series object reference so it never re-runs when series.ring is
mutated in place; change the dependency array to include a derived value
representing data availability (e.g., compute const pointsCount =
series.ring.bucket(5000).length and add pointsCount to the useEffect deps
alongside containerRef/series) so the effect re-runs when the ring buffer fills;
apply the same pattern to the other useEffect at the 169–171 region so both
effects respond to changes in the ring buffer point count rather than the series
object identity.
In `@dashboard/src/components/TraceExplorer.tsx`:
- Around line 844-845: The firstSeenRef Map (firstSeenRef) keeps growing because
entries are never removed as traces expire; update the code path that
updates/receives the current traces (the component logic that uses traces and
the block around lines ~861-873) to prune firstSeenRef by computing the current
set of trace_id values (from traces) and deleting any keys in firstSeenRef that
are not present, or alternatively delete entries when a trace is explicitly
removed/aged out; ensure this pruning runs whenever the traces array changes
(e.g., inside the effect or update function that processes traces) and consider
a periodic cleanup or cleanup on unmount if appropriate.
- Around line 44-126: TraceExplorer.tsx currently defines local duplicates of
computeStats and buildLanes; remove those local implementations and import the
tested helpers from dashboard/src/lib/stats (specifically computeStats and
buildLanes) and use them directly (retain SCAN_LANE and Lane/LanesResult types
only if not exported—otherwise import types too) so the runtime uses the same
logic as the unit-tested module; update any references in TraceExplorer to call
the imported computeStats and buildLanes and delete the duplicated functions.
- Around line 1038-1044: In TraceExplorer, the toggle button disappears after
expanding because the render guard uses hiddenCount > 0; change the conditional
to render when either there are hidden workers or the view is expanded (e.g.,
use (hiddenCount > 0 || showAll)) so the button stays visible when showAll is
true, and keep the existing onClick handler (setShowAll) and label logic (use
hiddenCount for the collapsed label, and the existing `showAll ? '▲ show fewer
workers' : '▼ show ${hiddenCount} more worker...'` text) so users can collapse
back.
- Around line 679-681: The click handling computes W, minMs and clickMs but
doesn't ignore clicks in the left label gutter or right PEND column; update the
click handler (the code that computes W, LABEL_W, PEND_BOX_W, minMs, clickMs) to
first compute drawableLeft = rect.left + LABEL_W and drawableRight =
drawableLeft + W and return early (or ignore the event) if clientX <
drawableLeft || clientX > drawableRight so clickMs is only calculated for clicks
inside the drawable chart area. Ensure you use the existing symbols W, LABEL_W,
PEND_BOX_W, minMs, clickMs and keep the projected time calculation unchanged
when inside bounds.
In `@dashboard/src/lib/ring.ts`:
- Around line 35-50: The bucket(bucketMs: number, maxAgeMs: number = 5 * 60 *
1000): DataPoint[] method uses bucketMs as a divisor and must validate it first;
add a guard at the start of bucket (before calling points()) that ensures
bucketMs is a finite positive number (e.g. Number.isFinite(bucketMs) && bucketMs
> 0) and handle invalid values by either throwing a clear error or returning an
empty array, so the subsequent Math.floor(p.t / bucketMs) and t + bucketMs / 2
calculations cannot produce NaN/Infinity; reference the bucket function,
bucketMs parameter, DataPoint type, and points() call when applying the fix.
In `@dashboard/src/style.css`:
- Around line 674-679: The .t2-legend-note rule uses var(--t5) which isn't
defined in :root, so add a definition for --t5 in the stylesheet's :root (or
replace var(--t5) with an existing token like var(--t4) if that was intended);
locate the :root block and either add e.g. --t5: <desired-color>; or update the
.t2-legend-note color to use an existing variable to ensure the rule is applied.
In `@dashboard/tsconfig.tsbuildinfo`:
- Line 1: The committed tsconfig.tsbuildinfo is a generated TypeScript build
cache and should be removed from version control: delete tsconfig.tsbuildinfo
from the commit and stop tracking it (git rm --cached tsconfig.tsbuildinfo or
remove from the index), then add tsconfig.tsbuildinfo to .gitignore to prevent
future commits; update the PR to only include the .gitignore change and the
removal of the tracked file.
---
Outside diff comments:
In `@crates/logfwd/src/worker_pool.rs`:
- Around line 476-515: queue_wait_ns is measured in the worker loop but never
propagated; add it to the acknowledgement path so downstream
pipeline/diagnostics see the worker-pool wait. Update the AckItem type to
include queue_wait_ns, populate that field in the worker code where
ack_tx.send(AckItem { ... }) is called (the block that calls process_item and
metrics.finish_active_batch), and ensure the pipeline/diagnostics consumers
(where AckItem is handled) read and serialize/attach queue_wait_ns to
spans/metrics (e.g. the code paths in pipeline.rs and diagnostics.rs that
currently record input-channel wait). Also consider setting the output span
field (send_ns/recv_ns or a new queue_wait field) via tracing when creating
output_span so traces reflect the worker-pool queue wait.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Repository YAML (base), Organization UI (inherited)
Review profile: ASSERTIVE
Plan: Pro
Run ID: c608a189-7686-41a6-add6-933b19bf7391
⛔ Files ignored due to path filters (2)
Cargo.lockis excluded by!**/*.lockdashboard/package-lock.jsonis excluded by!**/package-lock.json
📒 Files selected for processing (40)
.github/workflows/ci.ymlbench/scenarios/self-bench.yamlcrates/logfwd-bench/benches/pipeline.rscrates/logfwd-config/src/lib.rscrates/logfwd-core/benches/scanner.rscrates/logfwd-io/src/dashboard.htmlcrates/logfwd-io/src/diagnostics.rscrates/logfwd-output/Cargo.tomlcrates/logfwd-output/src/elasticsearch.rscrates/logfwd/src/pipeline.rscrates/logfwd/src/worker_pool.rsdashboard/.oxlintrc.jsondashboard/biome.jsondashboard/package.jsondashboard/src/api.tsdashboard/src/app.tsxdashboard/src/components/Chart.tsxdashboard/src/components/ConfigView.tsxdashboard/src/components/LogViewer.tsxdashboard/src/components/MetricBadges.tsxdashboard/src/components/PipelineView.tsxdashboard/src/components/StatusBar.tsxdashboard/src/components/TraceExplorer.tsxdashboard/src/lib/format.tsdashboard/src/lib/ring.tsdashboard/src/lib/stats.tsdashboard/src/style.cssdashboard/src/test/format.test.tsdashboard/src/test/lanes.test.tsdashboard/src/test/layout.test.tsdashboard/src/test/rates.test.tsdashboard/src/test/ring.test.tsdashboard/src/test/setup.tsdashboard/src/test/stats.test.tsdashboard/src/types.tsdashboard/tsconfig.app.jsondashboard/tsconfig.jsondashboard/tsconfig.tsbuildinfodashboard/vite.config.tsdashboard/vitest.config.ts
Comment on lines
+77
to
+80
| - name: run tsgo | ||
| if: matrix.check == 'typecheck' | ||
| working-directory: dashboard | ||
| run: npx tsgo --noEmit |
There was a problem hiding this comment.
🧹 Nitpick | 🔵 Trivial
🧩 Analysis chain
🏁 Script executed:
#!/bin/bash
set -euo pipefail
python - <<'PY'
import json, pathlib
pkg_path = pathlib.Path("dashboard/package.json")
lock_path = pathlib.Path("dashboard/package-lock.json")
pkg = json.loads(pkg_path.read_text())
deps = {}
deps.update(pkg.get("dependencies", {}))
deps.update(pkg.get("devDependencies", {}))
print("typecheck script:", pkg.get("scripts", {}).get("typecheck"))
print("tsgo version pinned in deps/devDeps:", deps.get("tsgo"))
if lock_path.exists():
lock_text = lock_path.read_text()
print("package-lock contains tsgo entry:", '"tsgo"' in lock_text)
else:
print("package-lock.json missing")
PYRepository: strawgate/memagent
Length of output: 174
Consider explicitly pinning tsgo in devDependencies for clarity.
The workflow runs npx tsgo --noEmit directly. While tsgo is already locked in package-lock.json (ensuring CI reproducibility), it's not explicitly listed in package.json. Adding it to devDependencies makes the dependency explicit and avoids potential surprises if the transitive dependency path changes.
Alternative: Run npm run typecheck instead (assuming the script is already defined) to keep intent explicit and tied to package.json.
🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed.
In @.github/workflows/ci.yml around lines 77 - 80, The workflow uses npx tsgo
--noEmit directly which relies on a transitive lockfile entry; add tsgo to
dashboard package.json devDependencies (e.g., "tsgo": "<pinned-version>") so the
tool is an explicit devDependency, or alternatively modify the CI step to run
the package script (e.g., npm run typecheck) that invokes tsgo from
package.json; update either the devDependencies entry for tsgo or the workflow
run line to npm run typecheck and ensure the package.json contains the
corresponding "typecheck" script.
| type: generator | ||
| transform: | | ||
| SELECT * FROM logs WHERE level_str != 'DEBUG' | ||
| SELECT * FROM logs WHERE level != 'DEBUG' |
There was a problem hiding this comment.
Use the benchmark-standard level_str column in the filter.
Line 4 diverges from existing benchmark/transform SQL usage (level_str) and can fail when the input schema doesn’t expose level.
Suggested fix
- SELECT * FROM logs WHERE level != 'DEBUG'
+ SELECT * FROM logs WHERE level_str != 'DEBUG'📝 Committable suggestion
‼️ IMPORTANT
Carefully review the code before committing. Ensure that it accurately replaces the highlighted code, contains no missing lines, and has no issues with indentation. Thoroughly test & benchmark the code to ensure it meets the requirements.
Suggested change
| SELECT * FROM logs WHERE level != 'DEBUG' | |
| SELECT * FROM logs WHERE level_str != 'DEBUG' |
🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed.
In `@bench/scenarios/self-bench.yaml` at line 4, The SQL in the benchmark scenario
uses a non-standard column "level" in the WHERE clause which can break when
input schemas lack that field; update the query in
bench/scenarios/self-bench.yaml to use the benchmark-standard column "level_str"
(e.g., replace "level != 'DEBUG'" with "level_str != 'DEBUG'") so it matches
other benchmark/transform SQL and avoids missing-column errors.
Comment on lines
+291
to
+296
| /// Maximum number of concurrent output workers. Default: 4. | ||
| pub workers: Option<usize>, | ||
| /// Batch target size in bytes before flushing. Default: 4 MiB. | ||
| pub batch_target_bytes: Option<usize>, | ||
| /// Batch flush timeout in milliseconds. Default: 100. | ||
| pub batch_timeout_ms: Option<u64>, |
There was a problem hiding this comment.
Validate workers at config-load time to prevent runtime panic.
workers: 0 currently parses, but later hits assert!(max_workers >= 1) in OutputWorkerPool::new, turning a user config error into a crash instead of a validation error.
Proposed fix
diff --git a/crates/logfwd-config/src/lib.rs b/crates/logfwd-config/src/lib.rs
@@
for (name, pipe) in &self.pipelines {
+ if let Some(workers) = pipe.workers
+ && workers == 0
+ {
+ return Err(ConfigError::Validation(format!(
+ "pipeline '{name}': workers must be >= 1"
+ )));
+ }
+
if pipe.inputs.is_empty() {
return Err(ConfigError::Validation(format!(
"pipeline '{name}' has no inputs"🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed.
In `@crates/logfwd-config/src/lib.rs` around lines 291 - 296, The config currently
allows workers: 0 which later triggers assert!(max_workers >= 1) in
OutputWorkerPool::new; update config validation when loading/parsing the struct
that contains the workers field to reject zero values (e.g., return a validation
error or map 0 to a meaningful default) so users get a clear config error
instead of a panic. Specifically, add a check for the workers field (pub
workers: Option<usize>) in the config-load/validate path and ensure Any Some(0)
is treated as invalid (with an Err containing a clear message) or normalized to
a minimum of 1 before passing to OutputWorkerPool::new. Ensure error text
references "workers" so callers can fix the config.
Comment on lines
+119
to
+132
| /// In-flight batch being processed right now. | ||
| pub struct ActiveBatch { | ||
| pub start_unix_ns: u64, | ||
| pub scan_ns: u64, | ||
| pub transform_ns: u64, | ||
| /// Current stage: "scan" | "transform" | "output" | ||
| pub stage: &'static str, | ||
| /// Unix ns when the current stage started (for frontend live duration) | ||
| pub stage_start_unix_ns: u64, | ||
| /// Worker id once assigned (-1 = not yet assigned / in queue) | ||
| pub worker_id: i64, | ||
| /// Unix ns when the worker actually started processing (0 = not yet) | ||
| pub output_start_unix_ns: u64, | ||
| } |
There was a problem hiding this comment.
🛠️ Refactor suggestion | 🟠 Major
Use typed absence instead of -1 / 0 sentinels.
worker_id: -1 and output_start_unix_ns: 0 are now part of the API contract and force the frontend to carry magic checks like worker_id < 0. Please model "not assigned yet" with Option instead, and preferably replace the stringly stage with an enum as well. As per coding guidelines: "No feature flags for behavior, no speculative traits, no sentinel values where Option applies".
🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed.
In `@crates/logfwd-io/src/diagnostics.rs` around lines 119 - 132, The ActiveBatch
struct uses sentinel values (worker_id = -1, output_start_unix_ns = 0) and a
stringly-typed stage; change worker_id to Option<i64> and output_start_unix_ns
to Option<u64> so absence is explicit, and replace stage: &'static str with a
Stage enum (e.g. Scan/Transform/Output) to avoid string comparisons; update
ActiveBatch, its constructor/serializers/deserializers, and any code that reads
these fields (consumers/tests) to handle Option and match on Stage instead of
using magic numeric/special-case checks.
- Expand inline return arrow functions to multi-line (app.tsx, LogViewer.tsx, TraceExplorer.tsx) to match biome formatter output - Lower vitest coverage thresholds from 70%/65% to 25%/20% to reflect that UI components are not yet unit-tested Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- pipeline.rs: validate workers >= 1 and batch_target_bytes > 0 at construction time, returning Err instead of panicking - elasticsearch.rs: skip whitespace after "took": colon in extract_took - ring.ts: guard bucketMs <= 0 / non-finite before division - app.tsx: memoize series slices (stable refs prevent chart churn on each poll) - TraceExplorer.tsx: keep show-fewer-workers button visible when showAll=true; prune firstSeenRef when traces age out - style.css: fix undefined --t5 variable reference → --t4 - .gitignore: exclude *.tsbuildinfo build artifacts; untrack tsconfig.tsbuildinfo Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
This was referenced Apr 10, 2026
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.
Summary
reqwest::Clientacross workers,filter_pathon_bulkURL, gzip fast (level 1) compression — ~2x throughput improvementfinish_active_batchnow called directly inworker_taskafterprocess_itemcompletes, instead of waiting for the pipeline ack select loop which was starved byflush_batch.awaitworkers,batch_target_bytes,batch_timeout_msnow configurable in YAMLlayoutSwimlane()extracted as pure function; scan bar clamped tonowMs(fixes bar glued to right edge on clock skew); Chart[series]dep added to plot useEffect (fixes chart reset on every parent render); 5s window option; window auto-selected from avgtotal_ns, locked on first user clickTest plan
just bench-selfand openhttp://127.0.0.1:9090— swimlane shows active batches without bars glued to right edgenpm testindashboard/→ 168/168 passing🤖 Generated with Claude Code