feat(PROF-15098): add batch reapply-by-id-and-bytes context API by jbachorik · Pull Request #598 · DataDog/java-profiler

jbachorik · 2026-06-15T15:06:25Z

What does this PR do?:

Adds setContextAttributesByIdAndBytes(int[], byte[][]) to ThreadContext, JavaProfiler, and ContextSetter. For each slot whose constantId > 0, writes both the DD sidecar encoding (read by the signal handler) and the OTEP attrs_data UTF-8 value (read by external eBPF/OTEP profilers) inside a single detach/attach window — no String allocation, hashing, or Dictionary lookup.

Also fixes the snapshotTags(int[]) guard from <= to >= (the old condition silently skipped the copy when the snapshot array was exactly the right size), and links the two independent MAX_CUSTOM_SLOTS / TAGS_STORAGE_LIMIT constants with cross-referencing comments.

Fixes ContextWallClockTest weight attribution: method1Impl accumulates wall-clock samples during monitor.wait() while method3 runs concurrently on the executor thread. Those samples were inflating method1's share to ~55% instead of the expected ~33%. Excluding Object.wait frames from the method1Impl attribution measures self-time only. The 0.3 allowedError for vm/vmx/fp/dwarf modes is retained to cover DWARF/VM trace-ID fragmentation that sheds some method2/method3 samples.

Motivation:

PROF-15098 — dd-trace-java's reapplyAppContext hot path must restore application-managed context attributes after a setContext span-activation wipes all custom slots. The existing setContextAttribute(int, String) API re-hashes and re-encodes the value on every call; the new API accepts the precomputed constant ID (from snapshotTags) and UTF-8 bytes (retained from the original String) so the hot path is zero-allocation.

Implementation notes:

Two-pass validation (check all slots before any write) prevents half-written records on malformed input.
setContextAttributesByIdAndBytes returns false and skips all writes when the record is already invalid (valid=0), preventing accidental resurrection of a cleared span-less record.
All N slot writes share one detach/attach window so a signal handler never observes a partially re-applied attribute set.
A shared writeSlot helper is used by both setContextAttributeDirect and the new batch path to avoid duplication of the sidecar + attrs_data write logic.
The (constantId, bytes) pairing is unverifiable by the callee; the contract is documented in the Javadoc.

Benchmarks (macOS/arm64, JDK 21, avgt ns/op):

Benchmark	Score	Error
`reapplyByIdAndBytes` — bare reapply, IDs+bytes already in hand	115 ± 1.5 ns
`reapplyCycle` — full cycle: `setContext` wipe + reapply (1t)	45 ± 0.8 ns
`reapplyCycle_2t`	46 ± 0.3 ns
`reapplyCycle_4t`	60 ± 1.4 ns
`spanLifecycle` (baseline: `setContext` + one `setContextAttribute`)	14 ns
`setContextFull`	6 ns

reapplyCycle at 45 ns restores 5 slots (compact + insert each) vs spanLifecycle at 14 ns for a single attribute — expected given the 5× attrs_data work. Scales cleanly under thread contention (46→60 ns at 4 threads, same pattern as spanLifecycle).

How to test the change?:

Six new tests in TagContextTest (deterministic readContextAttribute / snapshotTags readback — no flaky JFR-sample assertions):

testReapplyByIdAndBytes — happy path: set, snapshot, wipe via setContext, reapply, verify both sidecar and attrs_data views
testReapplyByIdAndBytesRejectsBadArgs — null arrays, length mismatch, null/oversized bytes
testReapplyByIdAndBytesReplacesExistingValue — compact-then-insert path (overwrites existing value)
testReapplyByIdAndBytesAfterClear — reapply after per-slot clearContextValue (valid stays 1)
testReapplyByIdAndBytesClearedRecord — reapply on a clearContext-cleared record (valid=0); must return false and not resurrect
testReapplyByIdAndBytesOverflowRollback — 10×255-byte values overflow attrs_data; overflowed slot's sidecar zeroed; slots written before the overflow are retained

Stress coverage: ReapplyContextAntagonist added to the chaos harness — drives the full reapply cycle from 8 threads while wall-clock signals fire, racing the detach/attach window. Run with --antagonists reapply-context.

For Datadog employees:

If this PR touches code that signs or publishes builds or packages, or handles
credentials of any kind, I've requested a security review (run the dd:platform-security-review
skill, or file a request via the PSEC review form).
bewaire also runs automatically on every PR.
This PR doesn't touch any of that.
JIRA: PROF-15098

Unsure? Have a question? Request a review!

Co-Authored-By: Claude Sonnet 4.6 (1M context) <noreply@anthropic.com>

datadog-datadog-prod-us1-2 · 2026-06-15T15:27:15Z

✨ Fix all issues with BitsAI

⚠️ Warnings

🚦 14 Pipeline jobs failed

DataDog/java-profiler | benchmarks-candidate-aarch64: [alloc]

DataDog/java-profiler | benchmarks-candidate-aarch64: [cpu,wall,alloc,memleak]

DataDog/java-profiler | benchmarks-candidate-aarch64: [cpu,wall]

View all 14 failed jobs.

Useful? React with 👍 / 👎

_{This comment will be updated automatically if new data arrives.

🔗 Commit SHA: 9f3e4ed | Docs | Datadog PR Page | Give us feedback!}

dd-octo-sts · 2026-06-15T15:32:41Z

CI Test Results

Run: #27773866793 | Commit: 264552f | Duration: 14m 56s (longest job)

✅ All 32 test jobs passed

Status Overview

JDK	glibc-aarch64/debug	glibc-amd64/debug	musl-aarch64/debug	musl-amd64/debug
8	-	✅	-	-
8-ibm	-	✅	-	-
8-j9	✅	✅	-	-
8-librca	-	-	✅	✅
8-orcl	-	✅	-	-
11	-	✅	-	-
11-j9	✅	✅	-	-
11-librca	-	-	✅	✅
17	✅	✅	-	-
17-graal	✅	✅	-	-
17-j9	✅	✅	-	-
17-librca	-	-	✅	✅
21	✅	✅	-	-
21-graal	✅	✅	-	-
21-librca	-	-	✅	✅
25	✅	✅	-	-
25-graal	✅	✅	-	-
25-librca	-	-	✅	✅

Legend: ✅ passed | ❌ failed | ⚪ skipped | 🚫 cancelled

Summary: Total: 32 | Passed: 32 | Failed: 0

Updated: 2026-06-18 16:48:54 UTC

Co-Authored-By: Claude Sonnet 4.6 (1M context) <noreply@anthropic.com>

- setContextAttributesByIdAndBytes: return false when constantIds.length > MAX_CUSTOM_SLOTS - snapshotTags(int[]): zero indices beyond attributes.size() to prevent foreign sidecar leakage - Add 5 acceptance tests

jbachorik · 2026-06-17T18:23:10Z

@codex review

Copilot

Pull request overview

This PR introduces a zero-allocation “reapply app context” path by adding a batch API that restores custom context attributes using precomputed (constant-id, UTF-8-bytes) pairs in a single detach/attach window, and fixes a snapshotTags(int[]) guard bug that previously skipped exact-sized buffers.

Changes:

Add setContextAttributesByIdAndBytes(int[], byte[][]) to ThreadContext, JavaProfiler, and ContextSetter, reusing a shared slot-write helper to keep DD sidecar + OTEP attrs_data consistent.
Fix ContextSetter.snapshotTags(int[]) to copy for exact-sized buffers and zero-out extra indices for oversized buffers.
Add targeted tests, plus JMH + chaos-harness coverage for the reapply cycle under signal pressure.

Reviewed changes

Copilot reviewed 8 out of 8 changed files in this pull request and generated 4 comments.

Show a summary per file

File	Description
ddprof-test/src/test/java/com/datadoghq/profiler/context/TagContextTest.java	Adds unit tests for batch reapply-by-id-and-bytes behavior and snapshotTags buffer-size semantics.
ddprof-lib/src/main/java/com/datadoghq/profiler/ThreadContext.java	Implements the batch reapply API and factors shared write logic to keep sidecar and `attrs_data` updates atomic.
ddprof-lib/src/main/java/com/datadoghq/profiler/JavaProfiler.java	Exposes the new batch API through the public profiler facade.
ddprof-lib/src/main/java/com/datadoghq/profiler/ContextSetter.java	Fixes snapshotTags guard/zeroing and adds a convenience wrapper for the new batch API.
ddprof-stresstest/src/jmh/java/com/datadoghq/profiler/stresstest/scenarios/throughput/ThreadContextBenchmark.java	Adds JMH benchmarks for the bare reapply and full reapply cycle under different thread counts.
ddprof-stresstest/src/chaos/java/com/datadoghq/profiler/chaos/ReapplyContextAntagonist.java	Adds a chaos antagonist to hammer the reapply cycle concurrently while wall-clock signals fire.
ddprof-stresstest/src/chaos/java/com/datadoghq/profiler/chaos/Main.java	Registers the new antagonist under `--antagonists reapply-context`.
ddprof-stresstest/build.gradle.kts	Adds `ddprof-lib` API compile-only dependency for the chaos source set.

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

chatgpt-codex-connector

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 00194f8ae3

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

Open a pull request for review
Mark a draft as ready
Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

jbachorik · 2026-06-17T20:22:30Z

@codex review

Copilot

Pull request overview

Copilot reviewed 8 out of 8 changed files in this pull request and generated 3 comments.

chatgpt-codex-connector

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: fc14d7cc36

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

Open a pull request for review
Mark a draft as ready
Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

…ClockTest method1Impl blocks in monitor.wait() while method3 runs on the executor thread. Those WAITING samples were counted as method1Weight, inflating it to ~55% instead of the expected ~33%. Exclude stack traces containing "Object.wait" from the method1 attribution so the assertion measures self-time only. Remove the 0.3 allowedError relaxation that was masking the root cause. Co-Authored-By: Claude Sonnet 4.6 (1M context) <noreply@anthropic.com>

Co-Authored-By: Claude Sonnet 4.6 (1M context) <noreply@anthropic.com>

The Object.wait exclusion fixed the method1Impl double-counting, but the 0.2 tolerance is too tight for DWARF/VM modes where native-frame PC variation fragments trace IDs and sheds method2/method3 samples. Co-Authored-By: Claude Sonnet 4.6 (1M context) <noreply@anthropic.com>

rkennke

Good change!
Have you verified that dd-trace-java is ok with the flip <= vs >= semantics in snapshotTags(int[])?

Oversized (length > size): old = silent no-op (the bug); new = copies correctly.
Undersized (length < size): old = partial copy up to length; new = no-op.

So any external caller that passed a deliberately-undersized buffer expecting a partial copy would now get nothing.

jbachorik · 2026-06-18T10:57:26Z

Have you verified that dd-trace-java is ok with the flip

There will be a followup PR for dd-trace-java - it will need to get more adjustments for this functionality, anyhow

ivoanjo

👍 This looks good to me. I've left a bunch of comments, but they're all in the "we can remove a bit of code here, we can throw instead of swallowing failure here, we can simplify a bit here" vibe so nothing big

- Remove snapshotTags(int[]) overload; inline copyTags in no-arg form - Make MAX_CUSTOM_SLOTS package-private; drop duplicate TAGS_STORAGE_LIMIT - Throw NPE/IAE for null/mismatched args in setContextAttributesByIdAndBytes - Remove pre-validation loop; let write loop handle bad utf8 naturally - Drop "Delegates to" impl detail from JavaProfiler Javadoc - Use distinct traceIdLow (not spanId) in antagonist, benchmark, tests - Delete 3 dead snapshotTags(int[]) sizing tests Co-Authored-By: Claude Sonnet 4.6 (1M context) <noreply@anthropic.com>

snapshotTags(int[]) is used in dd-trace-java PR 11646 — removing it broke the caller. Restore the overload and the 3 contract tests. Co-Authored-By: Claude Sonnet 4.6 (1M context) <noreply@anthropic.com>

dd-octo-sts · 2026-06-18T13:25:38Z

Reliability & Chaos Results

✅ All reliability & chaos checks passed Pipeline: https://gitlab.ddbuild.io/DataDog/java-profiler/-/pipelines/119752144

Move null + length checks for active slots into a pre-pass before detach() so a thrown NPE/IAE never leaves the record with valid=0. Add post-throw sidecar assertions to the test to catch regressions. Add int/uint32 safety comment in sframe.cpp. Co-Authored-By: Claude Sonnet 4.6 (1M context) <noreply@anthropic.com>

- ReapplyContextAntagonist: throw ISE instead of counting reapply failures — reapply must always succeed when record is valid - Remove now-dead reapplyFailures field and its stopGracefully logging - TagContextTest: hoist Assumptions.assumeTrue(!Platform.isJ9()) into @beforeeach to avoid repeating it in every test method Co-Authored-By: Claude Sonnet 4.6 (1M context) <noreply@anthropic.com>

ivoanjo · 2026-06-18T17:34:45Z

+        // the record detached (valid=0) after an exception unwinds past attach().
+        for (int i = 0; i < len; i++) {
+            if (constantIds[i] > 0) {
+                byte[] bytes = Objects.requireNonNull(utf8[i], "utf8[" + i + "]");


Wait, isn't this an allocation? 👀

Specifically

"utf8[" + i + "]"

...?

Yes 😱

#606

feat(PROF-15098): add batch reapply-by-id-and-bytes context API

13a9fdc

Co-Authored-By: Claude Sonnet 4.6 (1M context) <noreply@anthropic.com>

jbachorik added the AI label Jun 15, 2026

test(PROF-15098): add reapply benchmark and chaos antagonist

a38d084

Co-Authored-By: Claude Sonnet 4.6 (1M context) <noreply@anthropic.com>

jbachorik and others added 4 commits June 15, 2026 17:41

fix(bench): stabilize reapplyByIdAndBytes with Level.Iteration reset

4d26a77

Co-Authored-By: Claude Sonnet 4.6 (1M context) <noreply@anthropic.com>

Merge branch 'main' into prof-15098-context-by-id

6479338

fix(review): address sphinx review findings

9ac8760

Co-Authored-By: Claude Sonnet 4.6 (1M context) <noreply@anthropic.com>

Merge branch 'main' into prof-15098-context-by-id

97e40c8

jbachorik commented Jun 16, 2026

View reviewed changes

Comment thread ddprof-lib/src/main/java/com/datadoghq/profiler/ThreadContext.java Outdated

Comment thread ddprof-lib/src/main/java/com/datadoghq/profiler/ContextSetter.java

fix: address review feedback on PR #598

00194f8

- setContextAttributesByIdAndBytes: return false when constantIds.length > MAX_CUSTOM_SLOTS - snapshotTags(int[]): zero indices beyond attributes.size() to prevent foreign sidecar leakage - Add 5 acceptance tests

jbachorik requested a review from Copilot June 17, 2026 18:23

Copilot started reviewing on behalf of jbachorik June 17, 2026 18:23 View session

Copilot AI reviewed Jun 17, 2026

View reviewed changes

chatgpt-codex-connector Bot reviewed Jun 17, 2026

View reviewed changes

Comment thread ddprof-lib/src/main/java/com/datadoghq/profiler/JavaProfiler.java

Comment thread ddprof-lib/src/main/java/com/datadoghq/profiler/ContextSetter.java

jbachorik added 2 commits June 17, 2026 21:43

address PR #598 review feedback

3d36764

Merge branch 'main' into prof-15098-context-by-id

fc14d7c

jbachorik requested a review from Copilot June 17, 2026 20:22

Copilot started reviewing on behalf of jbachorik June 17, 2026 20:22 View session

Copilot AI reviewed Jun 17, 2026

View reviewed changes

Comment thread ddprof-lib/src/main/java/com/datadoghq/profiler/JavaProfiler.java Outdated

Comment thread ddprof-lib/src/main/java/com/datadoghq/profiler/JavaProfiler.java Outdated

Comment thread ddprof-lib/src/main/java/com/datadoghq/profiler/JavaProfiler.java Outdated

chatgpt-codex-connector Bot reviewed Jun 17, 2026

View reviewed changes

Comment thread ddprof-lib/src/main/java/com/datadoghq/profiler/JavaProfiler.java Outdated

jbachorik and others added 3 commits June 17, 2026 22:52

fix(test): correct comment on Object.wait exclusion

9f118b8

Co-Authored-By: Claude Sonnet 4.6 (1M context) <noreply@anthropic.com>

jbachorik marked this pull request as ready for review June 17, 2026 21:45

jbachorik requested a review from a team as a code owner June 17, 2026 21:45

jbachorik added the AI: Author Signed-off label Jun 18, 2026

jbachorik requested a review from ivoanjo June 18, 2026 09:03

rkennke approved these changes Jun 18, 2026

View reviewed changes

ivoanjo approved these changes Jun 18, 2026

View reviewed changes

jbachorik and others added 2 commits June 18, 2026 14:44

restore snapshotTags(int[]) overload and sizing tests

02e1b02

snapshotTags(int[]) is used in dd-trace-java PR 11646 — removing it broke the caller. Restore the overload and the 3 contract tests. Co-Authored-By: Claude Sonnet 4.6 (1M context) <noreply@anthropic.com>

jbachorik and others added 7 commits June 18, 2026 15:40

fix test: expect IAE instead of false for oversized constantIds array

f08ff92

fix: throw for null/oversized utf8 in write loop; update tests

48ba127

add comment explaining J9 skip in TagContextTest

a8bcb27

remove unjustified CPU spin from ReapplyContextAntagonist

999a704

fix incoherent Javadoc in ReapplyContextAntagonist

9f3e4ed

jbachorik merged commit bbd314b into main Jun 18, 2026
112 checks passed

jbachorik deleted the prof-15098-context-by-id branch June 18, 2026 17:20

github-actions Bot added this to the 1.45.0 milestone Jun 18, 2026

ivoanjo reviewed Jun 18, 2026

View reviewed changes

Conversation

jbachorik commented Jun 15, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

datadog-datadog-prod-us1-2 Bot commented Jun 15, 2026 • edited by datadog-prod-us1-4 Bot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

⚠️ Warnings

Uh oh!

dd-octo-sts Bot commented Jun 15, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

CI Test Results

Status Overview

Uh oh!

Uh oh!

Uh oh!

jbachorik commented Jun 17, 2026

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

chatgpt-codex-connector Bot left a comment

Choose a reason for hiding this comment

💡 Codex Review

Uh oh!

Uh oh!

Uh oh!

jbachorik commented Jun 17, 2026

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Uh oh!

Uh oh!

Uh oh!

Uh oh!

chatgpt-codex-connector Bot left a comment

Choose a reason for hiding this comment

💡 Codex Review

Uh oh!

Uh oh!

rkennke left a comment

Choose a reason for hiding this comment

Uh oh!

jbachorik commented Jun 18, 2026

Uh oh!

ivoanjo left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

dd-octo-sts Bot commented Jun 18, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Reliability & Chaos Results

Uh oh!

Uh oh!

ivoanjo Jun 18, 2026

Choose a reason for hiding this comment

Uh oh!

ivoanjo Jun 18, 2026

Choose a reason for hiding this comment

Uh oh!

jbachorik Jun 18, 2026

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

jbachorik commented Jun 15, 2026 •

edited

Loading

datadog-datadog-prod-us1-2 Bot commented Jun 15, 2026 •

edited by datadog-prod-us1-4 Bot

Loading

dd-octo-sts Bot commented Jun 15, 2026 •

edited

Loading

dd-octo-sts Bot commented Jun 18, 2026 •

edited

Loading