fix: Resolve race condition in check reducers#2947
fix: Resolve race condition in check reducers#2947adriantam merged 2 commits intoopenfga:mainopenfga/openfga:mainfrom rafanaskin:fix-request-cancelledrafanaskin/openfga:fix-request-cancelledCopy head branch name to clipboard
Conversation
WalkthroughThis PR adds early context cancellation checks to three check reducers (union, intersection, exclusion) to prevent non-deterministic nested handler execution when the parent context is cancelled, addressing a race condition bug. Changes
Estimated code review effort🎯 2 (Simple) | ⏱️ ~12 minutes Suggested reviewers
🚥 Pre-merge checks | ✅ 3✅ Passed checks (3 passed)
✏️ Tip: You can configure your own custom pre-merge checks in the settings. ✨ Finishing Touches🧪 Generate unit tests (beta)
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
6840500 to
9914ae1
Compare
Codecov Report✅ All modified and coverable lines are covered by tests. Additional details and impacted files@@ Coverage Diff @@
## main #2947 +/- ##
==========================================
+ Coverage 90.08% 90.20% +0.13%
==========================================
Files 186 186
Lines 19774 19808 +34
==========================================
+ Hits 17811 17866 +55
+ Misses 1310 1289 -21
Partials 653 653 ☔ View full report in Codecov by Sentry. 🚀 New features to boost your workflow:
|
cdcb928 to
848e4db
Compare
7b7cb72 to
4842f9d
Compare
There was a problem hiding this comment.
Pull request overview
This PR fixes a race condition in the check reducer functions (union, intersection, exclusion) in internal/graph/check.go. When a parent context was already cancelled before a reducer was invoked, the reducers would proceed to create goroutines and enqueue work, which could non-deterministically result in a Request Cancelled error rather than a correct check result. The fix adds a lightweight ctx.Err() != nil guard at the entry of each reducer, mirroring the existing guard already present in ResolveCheck.
Changes:
- Added early-exit
ctx.Err()guard tounion,intersection, andexclusionreducers to prevent goroutine launch on pre-cancelled context. - Added a CHANGELOG entry for the fix under
[Unreleased].
Reviewed changes
Copilot reviewed 2 out of 2 changed files in this pull request and generated 2 comments.
| File | Description |
|---|---|
internal/graph/check.go |
Adds ctx.Err() != nil pre-checks to union, intersection, and exclusion to short-circuit before goroutine creation when the parent context is already cancelled. |
CHANGELOG.md |
Adds a Fixed entry for the race condition in check reducers. |
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
There was a problem hiding this comment.
🧹 Nitpick comments (1)
CHANGELOG.md (1)
12-12: Add PR reference link for consistency.This entry is clear, but adding
[#2947](https://github.com/openfga/openfga/pull/2947)would match the surrounding changelog style and improve traceability.🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In `@CHANGELOG.md` at line 12, Update the changelog entry text to include the PR reference link for consistency by appending or replacing the sentence with the PR reference `[`#2947`](https://github.com/openfga/openfga/pull/2947)` so the line reads like the surrounding entries (e.g., "Fixed a race condition in check reducers causing non-deterministic nested handler execution due to canceled parent context. [`#2947`](https://github.com/openfga/openfga/pull/2947)").
🤖 Prompt for all review comments with AI agents
Verify each finding against the current code and only fix it if needed.
Nitpick comments:
In `@CHANGELOG.md`:
- Line 12: Update the changelog entry text to include the PR reference link for
consistency by appending or replacing the sentence with the PR reference
`[`#2947`](https://github.com/openfga/openfga/pull/2947)` so the line reads like
the surrounding entries (e.g., "Fixed a race condition in check reducers causing
non-deterministic nested handler execution due to canceled parent context.
[`#2947`](https://github.com/openfga/openfga/pull/2947)").
Added a `ctx.Err() != nil` guard to the check reducers (union, intersection, exclusion) to detect parent context cancellation before executing nested handlers. Depending on goroutine scheduling, this could cause the `Check` call to return `Request Cancelled` instead of `Allowed: true`.
4842f9d to
d3504c1
Compare
Add support for exporting logs via OTLP, allowing integration with any OpenTelemetry-compatible backend (Grafana Loki, Datadog, GCP Cloud Logging, etc.). When log.otlp.enabled is set, logs are exported via OTLP in addition to stdout. The otelzap bridge attaches span context (trace ID, span ID) to each log record, enabling log-trace correlation in backends that support it. Configuration mirrors the trace configuration shape: the standard OTel env vars (OTEL_EXPORTER_OTLP_LOGS_ENDPOINT, OTEL_EXPORTER_OTLP_ENDPOINT) only answer where logs are sent, while the explicit OpenFGA flag (log.otlp.enabled / --log-otlp-enabled / OPENFGA_LOG_OTLP_ENABLED) answers whether to export at all — so a cluster that injects the generic endpoint variable for tracing does not silently start exporting logs on upgrade. zap's production sampling is applied outside the tee, so the keep/drop decision is made once, before fan-out: stdout and OTLP receive the same deterministically sampled stream and collector egress stays bounded during log storms. Changes: - Add otelzap bridge core backed by an OTLP log provider (internal/telemetry/logging.go) - Add WithOTELCore logger option that tees log entries to the bridge. The bridge core is capped to the configured log level, the stdout core strips the bridge-only context field via contextFilterCore, and the sampler wraps the tee (pkg/logger/logger.go) - Attach the context field in the *WithContext logging methods only when an OTEL core is configured, and use those methods in the gRPC logging interceptor so the bridge can extract span context (pkg/middleware/logging/logging.go) - Add config: log.otlp.enabled, log.otlp.endpoint, log.otlp.tls.enabled with OPENFGA_LOG_OTLP_* env vars and OTEL_EXPORTER_OTLP_* endpoint fallbacks (pkg/server/config/config.go, cmd/run/, .config-schema.json) feat: add potential v2 breaking change logs for Expand and ListUsers (openfga#3182) release: update changelog for release 1.18.1 (openfga#3188) release: Update changelog to prep for 1.18.1 release (openfga#3186) test: fix flaky TestV2CheckWithIteratorCache_HigherConsistencyBypassesCache (openfga#3061) Co-authored-by: Joshua Jones <joshua.jones.software@gmail.com> Merge commit from fork * test reproducing ListUsers report * implement fix * additional test with 3 operands fix: match IPv4-mapped IPv6 addresses in the in_cidr condition (openfga#3181) Signed-off-by: kanywst <niwatakuma@icloud.com> fix: use deterministic proto marshaling for stored authorization models (openfga#3171) chore: create draft release and publish after provenance succeeds for immutable tags/releases (openfga#3178) docs: fix changelog entry (openfga#3177) Signed-off-by: Saad Hussain <saad.hussain@okta.com> feat: add v2Check logs for resolution breaking change (openfga#3149) feat: BatchCheck uses v2Check when ExperimentalWeightedGraphCheck is enabled (openfga#3154) Signed-off-by: Saad Hussain <saad.hussain@okta.com> chore: update changelog to add CVE identifiers for recent fixes (openfga#3176) chore: update todo comment string in migration guide (openfga#3175) release: update changelog for release `v1.18.0` (openfga#3174) Co-authored-by: adriantam <adrian.tam@okta.com> Merge commit from fork * fix(authn): require issuer and audience when OIDC authn is enabled Enforce configuration authn.oidc.audience and authn.oidc.issuer when `authn.method` is set to `oidc`. * fixed based on code review feedback * fix: adding comments + test case on empty space for audience * update based on code review feedback * Update CHANGELOG.md Co-authored-by: Joshua Jones <joshua.jones.software@gmail.com> * update changelog * Update CHANGELOG.md --------- Co-authored-by: Joshua Jones <joshua.jones.software@gmail.com> Merge commit from fork * fix(mysql): collate identifier columns as utf8mb4_bin * fix: add operator note to changelog * fix: move tests to shared storage * fix: separate migrations for each table * fix: add runbook for migrations * fix: set lock_wait_timeout * fix: add docker instructions * fix: add docker instructions * fix: combine the migrations back into one, fix documentation for this * fix: include details about table copy and * Update CHANGELOG.md Co-authored-by: Adrian Tam <adrian.tam@okta.com> --------- Co-authored-by: Adrian Tam <adrian.tam@okta.com> fix: use constant-time comparison for preshared key authentication (openfga#3168) chore(deps): bump the dependencies group with 2 updates (openfga#3166) Signed-off-by: dependabot[bot] <support@github.com> Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com> Co-authored-by: Justin Cohen <justincoh@gmail.com> chore(deps): bump the dependencies group with 2 updates (openfga#3167) Signed-off-by: dependabot[bot] <support@github.com> Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com> release: update changelog for release `v1.17.1` (openfga#3165) Signed-off-by: Saad Hussain <saad.hussain@okta.com> chore: bump grpc-health-probe to v0.4.52 (openfga#3164) Co-authored-by: Saad Hussain <saad.hussain@okta.com> docs: update caching docs (openfga#3163) Signed-off-by: Saad Hussain <saad.hussain@okta.com> fix: continuation token deserializer - handle `|` in type names (openfga#3152) chore(deps): bump the dependencies group across 1 directory with 13 updates (openfga#3162) Signed-off-by: dependabot[bot] <support@github.com> Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com> Co-authored-by: Justin Cohen <justincoh@gmail.com> chore(deps): bump the dependencies group across 1 directory with 10 updates (openfga#3156) Signed-off-by: dependabot[bot] <support@github.com> Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com> Co-authored-by: Justin Cohen <justincoh@gmail.com> fix: use query start time as iterator cache entry LastModified to prevent stale-read survival (openfga#3155) chore(deps): bump grpc-ecosystem/grpc-health-probe from v0.4.50 to v0.4.51 in the dependencies group (openfga#3157) Signed-off-by: dependabot[bot] <support@github.com> Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com> Co-authored-by: Saad Hussain <saad.hussain@okta.com> Co-authored-by: Justin Cohen <justincoh@gmail.com> chore: Bump go toolchain to 1.26.4 (openfga#3159) fix: prevent v2Check from falling back for throttling and validation (openfga#3150) ci: make pr benchmark comparisons less fragile (openfga#3153) Co-authored-by: coderabbitai[bot] <136622811+coderabbitai[bot]@users.noreply.github.com> release: update changelog for release `v1.17.0` (openfga#3151) Co-authored-by: Vic-Dev <vmichellej@gmail.com> refactor: redesign cache key generation, making it more secure and consistent (openfga#3148) feat: add configurable trace sampler with ParentBased support (openfga#3072) release: update changelog for release `v1.16.1` (openfga#3147) chore: update grpc-health-probe to latest to addres std lib CVEs (openfga#3146) fix: skip v2Check weight2 pruning if iterator is unordered (openfga#3145) Signed-off-by: Saad Hussain <saad.hussain@okta.com> chore: add concurrency-group to PR-related CI steps (openfga#3140) docs: move OIDC JWKS refresh entry from v1.15.1 to v1.16.0 (openfga#3142) fix: when v2Check is primary algorithm, fix fallback condition and emit metrics (openfga#3141) release: update changelog for release `v1.16.0` (openfga#3139) Merge commit from fork * fix: Standardize cache key generation everywhere * Revert "fix: Standardize cache key generation everywhere" This reverts commit ce60c6f9ae48a310a4dea2b824b5993520c8ba1c. * length-encoded approach * Revert "length-encoded approach" This reverts commit b94637c187c57a2d3f3904247267bbc8ad21a41a. * implement Hexer() and use in some cache keys * appendConditionsHash -> generateConditionsHash * mv Hexer -> BuildKey * BuildKey -> BuildCacheKey refactor to BuildCacheKey on each individual prefix component to remove collision risk * make v2 cacheKey functions the standard * make cache key prefixes consistent * generateConditionsHash -> generateConditionsString * delimit condition name string * remove extra empty check * relocate and reuse existing userTypeRestrictions function * replace hashing of conditions for brevity * fix tests * cleanup, add test file * remove unused slice capacity * make prefix treatments consistent * add nil-byte separator in object ids * hash user type restrictions for brevity * update comment string, use nil-byte separator for future-proofing * remove call to xxhash.new * adjustments for performance, consistency, and interface ergonomics. * make condition hash generation more efficient. * add clarifying comments for size caluculations * fix builder growth calculations to match original intentions * update outdated comment * don't hash potentially zero values * Revert "don't hash potentially zero values" This reverts commit 2ad6c5b796bc20a82c1414c5146c3c708330eefa. * add condition key separator unconditionally * patch test expectations for new key structure * export V2IteratorCachePrefix, use in tests --------- Co-authored-by: justin <justin.cohen@okta.com> Co-authored-by: Joshua Jones <joshua.jones@okta.com> fix: unintentional zeroing of slice values by setting slice to nil (openfga#3135) increase the check v2 trace information fidelity (openfga#3134) feat: add datastore ping and ping retry configurations (openfga#3113) fix: prevent v2Check strategies returning spurious false on context cancellation (openfga#3128) Signed-off-by: Saad Hussain <saad.hussain@okta.com> fix(authn/oidc): refresh JWKS on unknown kid to handle key rotation (openfga#3101) fix: v2Check falls back to v1 on errors (openfga#3126) fix: don't cache false results from cancelled-context goroutines in v2Check (openfga#3125) Signed-off-by: Saad Hussain <saad.hussain@okta.com> make union cache key unique by including the node input's label (openfga#3117) Signed-off-by: Saad Hussain <saad.hussain@okta.com> Co-authored-by: Saad Hussain <saad.hussain@okta.com> fix: shadow v2check and check use the same trace (openfga#3118) fix: bump go toolchain version to 1.26.3 (openfga#3115) chore: add more spans/attributes to v1 and v2 Check (openfga#3116) Signed-off-by: Saad Hussain <saad.hussain@okta.com> fix: add matches attribute to shadowv2Check and Check spans (openfga#3114) fix: use the same `request_id` for shadow traces in v2Check (openfga#3110) release: update changelog for release `v1.15.1` (openfga#3112) feat: reuse MySQL container across tests (openfga#3042) fix: close all channels opened thus far, before return on error (openfga#3111) chore: Add more spans/attributes to v2Check (openfga#3102) Signed-off-by: Saad Hussain <saad.hussain@okta.com> fix: ensure acquired limiter token is released on throttle context cancelation (openfga#3106) fix: cancel context before waiting on worker pool in ResolveUnionEdges (openfga#3105) fix: replace golang.org/x/exp/maps with stdlib maps to resolve govet inline errors (openfga#3104) fix: v2Check EdgeCacheKey collisions (openfga#3097) Signed-off-by: Saad Hussain <saad.hussain@okta.com> fix: expose context errors when they can be the potential cause of an underlying datastore error (openfga#3096) chore(deps): bump the dependencies group across 1 directory with 2 updates (openfga#3093) Signed-off-by: dependabot[bot] <support@github.com> Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com> fix: move subproblem cache lookup from ResolveUnionEdges into ResolveUnion (openfga#3095) chore(deps): bump the dependencies group with 4 updates (openfga#3092) Signed-off-by: dependabot[bot] <support@github.com> Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com> fix: error handler panic (openfga#3091) release: update changelog for release `v1.15.0` (openfga#3090) chore(deps): bump the dependencies group across 1 directory with 4 updates (openfga#3087) Signed-off-by: dependabot[bot] <support@github.com> Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com> fix: v2Check correctly uses query cache even when cache controller is disabled (openfga#3086) Signed-off-by: Saad Hussain <saad.hussain@okta.com> chore(deps): bump the dependencies group across 1 directory with 7 updates (openfga#3081) Signed-off-by: dependabot[bot] <support@github.com> chore(deps): bump github.com/jackc/pgx/v5 from 5.9.1 to 5.9.2 (openfga#3085) Signed-off-by: dependabot[bot] <support@github.com> chore(deps): bump grpc-ecosystem/grpc-health-probe from v0.4.47 to v0.4.48 in the dependencies group across 1 directory (openfga#3065) Signed-off-by: dependabot[bot] <support@github.com> feat: try to use UDS internally between HTTP server and gRPC server (openfga#2937) feat: add jitter to internal cache TTLs to prevent thundering herd effects (openfga#3033) Signed-off-by: Asish Kumar <officialasishkumar@gmail.com> list objects pipeline edge pruning (openfga#3075) update toolchain go to 1.26.2 to address stdlib CVEs (openfga#3084) chore: add store ID and datastore query/item count to shadowV2Check log (openfga#3073) Signed-off-by: Saad Hussain <saad.hussain@okta.com> CI speed up (openfga#3062) Co-authored-by: Joshua Jones <joshua.jones.software@gmail.com> Co-authored-by: Joshua Jones <joshua.jones@okta.com> Add tracing to v2 Check planner strategy selection (openfga#3077) fix: v2Check honours `check-query-cache-enabled` flag (openfga#3070) Signed-off-by: Saad Hussain <saad.hussain@okta.com> feat: reuse PostgreSQL container across tests (openfga#3018) chore: fix reporter links in changelog (openfga#3069) release: update changelog for release `v1.14.2` (openfga#3068) fix: add null byte delimiter in contextual tuple cache keys and validation in v2Check (openfga#3064) release: update changelog for release `v1.14.1` (openfga#3060) chore(deps): bump the dependencies group across 1 directory with 7 updates (openfga#3056) Signed-off-by: dependabot[bot] <support@github.com> Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com> Co-authored-by: Adrian Tam <adrian.tam@okta.com> fix: use a baseURL for AuthZEN configuration endpoint (openfga#3057) chore: replace docker with moby (openfga#3047) Iterator Cache V2: Storage Wrapper Pattern with Lock-Free Design (openfga#3016) Signed-off-by: Saad Hussain <saad.hussain@okta.com> Co-authored-by: Saad Hussain <saad.hussain@okta.com> Co-authored-by: Joshua Jones <joshua.jones.software@gmail.com> test: fix flaky condition test (openfga#3058) Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com> chore(deps): bump the dependencies group across 1 directory with 6 updates (openfga#3045) Signed-off-by: dependabot[bot] <support@github.com> Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com> chore(deps): bump go.opentelemetry.io/otel/exporters/otlp/otlptrace/otlptracehttp from 1.41.0 to 1.43.0 (openfga#3054) Signed-off-by: dependabot[bot] <support@github.com> Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com> Co-authored-by: Adrian Tam <adrian.tam@okta.com> feat: add graceful shutdown timeout configuration (openfga#2976) chore: update changelog to reflect playground off-by-default behavior (openfga#3053) chore: decouple Error() and Unwrap() (openfga#3051) perf: remove `fmt` from cache key generation (openfga#3006) perf: enhancements for list objects (openfga#3043) chore: remove stale notice file (openfga#3050) docs: update 1.14.0 for CVE fix (openfga#3046) feat: Add datastore throttling & concurrency limiting to v2Check (openfga#3035) Signed-off-by: Saad Hussain <saad.hussain@okta.com> release: update changelog for release `v1.14.0` (openfga#3040) batch check cache (openfga#3025) Merge commit from fork Also prevent enabling playground when the server requires authentication feat: add stats on tuple iterator query (openfga#3030) fix: remove unnecessary non-deterministic test (openfga#3038) remove unnecessary import (openfga#3032) perf: improve the intersection algorithm, reducing latency and memory use (openfga#3031) fix: ListObjects pipeline algorithm enhancements and fix for potential deadlock (openfga#3028) chore: Also update openfga/helm-charts in release script (openfga#3010) chore: update CICD to enforce GRPC healthprobe changes (openfga#2990) fix: SQL `TupleOperation` serialization and `pgx.ErrNoRows` error handling (openfga#3014) docs: update changelog for CVE-2026-33729 (openfga#3017) chore: output the diff after running keep-a-changelog (openfga#3015) chore(deps): bump the dependencies group across 1 directory with 11 updates (openfga#2998) Signed-off-by: dependabot[bot] <support@github.com> Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com> chore(deps): bump the dependencies group across 1 directory with 10 updates (openfga#2999) Signed-off-by: dependabot[bot] <support@github.com> Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com> release: update changelog for release `v1.13.1` (openfga#3002) Merge commit from fork * strip unicode control characters, encode context key length in cache key * add cache_key tests to ensure we escape control characters and encode context key length * run linter * docs: update changelog * Apply suggestion from @adriantam Co-authored-by: Adrian Tam <adrian.tam@okta.com> * Apply suggestion from @adriantam Co-authored-by: Adrian Tam <adrian.tam@okta.com> --------- Co-authored-by: Saad Hussain <saad.hussain@okta.com> Co-authored-by: Saad Hussain <saad.h@outlook.com> Co-authored-by: Adrian Tam <adrian.tam@okta.com> release: update changelog for release `v1.13.0` (openfga#2997) refactor: Separate caches for v1 and v2 Check (openfga#2968) Co-authored-by: Justin Cohen <justincoh@gmail.com> docs: fix typos in comments (openfga#2972) Signed-off-by: Artem Muterko <artem@sopho.tech> Co-authored-by: Adrian Tam <adrian.tam@okta.com> observability: aggregate message statistics for each list-objects sender into a single span (openfga#2993) Co-authored-by: Justin Cohen <justincoh@gmail.com> fix: capture panics in pipeline's base resolver, and return as errors. (openfga#2994) docs: fix typos in RELEASES.md and Makefile (openfga#2980) Co-authored-by: Adrian Tam <adrian.tam@okta.com> AuthZen v1.0 Implementation (openfga#2875) Signed-off-by: dependabot[bot] <support@github.com> Signed-off-by: Rokibul Hasan <mdrokibulhasan@appscode.com> Signed-off-by: Vihang Mehta <vihang@gimletlabs.ai> Co-authored-by: Karl Persson <kalle.persson92@gmail.com> Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com> Co-authored-by: Maria Ines Parnisari <maria.inesparnisari@okta.com> Co-authored-by: Rokibul Hasan <mdrokibulhasan18@gmail.com> Co-authored-by: José Padilla <jose.padilla@okta.com> Co-authored-by: Adrian Tam <adrian.tam@okta.com> Co-authored-by: Vihang Mehta <vihang@gimletlabs.ai> Co-authored-by: Joshua Jones <joshua.jones.software@gmail.com> Co-authored-by: Justin Cohen <justincoh@gmail.com> release: update changelog for release `v1.12.1` (openfga#2992) chore(deps): bump google.golang.org/grpc from 1.79.1 to 1.79.3 (openfga#2988) Signed-off-by: dependabot[bot] <support@github.com> Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com> Co-authored-by: Adrian Tam <adrian.tam@okta.com> chore: enforce minor version upgrade rule (openfga#2978) Co-authored-by: Copilot Autofix powered by AI <175728472+Copilot@users.noreply.github.com> perf: tuple validation performance tweaks (openfga#2984) fix: support URI schemes in OTLP endpoint configuration (openfga#2981) refactor: remove custom pipes and replace with channels (openfga#2977) chore(docs): added caching docs (openfga#2664) Co-authored-by: Saad Hussain <saad.hussain@okta.com> release: update changelog for release `v1.12.0` (openfga#2974) chore: update toolchain go to 1.26.1 (openfga#2975) fix: update toolchain go to 1.25.8 to address stdlib CVEs (openfga#2971) fix: correct swapped format args in DecodeParameterType error message (openfga#2961) Signed-off-by: Artem Muterko <artem@sopho.tech> Co-authored-by: Adrian Tam <adrian.tam@okta.com> perf: small tweaks to tuple validation functions (openfga#2963) Co-authored-by: Vic-Dev <vmichellej@gmail.com> test: add tests for condition parameter type any with complex context structures (openfga#2959) Signed-off-by: Artem Muterko <artem@sopho.tech> test: add functional test for ReadAssertions API endpoint (openfga#2960) Signed-off-by: Artem Muterko <artem@sopho.tech> Co-authored-by: Adrian Tam <adrian.tam@okta.com> make configurable the maximum received grpc message size (openfga#2952) Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com> fix: set ExperimentalPipelineListObjects in experimentals by default (openfga#2957) fix: `cache_item_count` metric overcounting (openfga#2950) Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com> chore(deps): bump the dependencies group across 1 directory with 7 updates (openfga#2953) Signed-off-by: dependabot[bot] <support@github.com> Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com> Co-authored-by: Adrian Tam <adrian.tam@okta.com> refactor: Make shadowV2Check final log clearer (openfga#2946) Move telemetry package to internal/telemetry (openfga#2938) Signed-off-by: Oleksandr Shestopal <ar.shestopal-oshegithub@gmail.com> Co-authored-by: Adrian Tam <adrian.tam@okta.com> chore(deps): bump the dependencies group across 1 directory with 3 updates (openfga#2956) Signed-off-by: dependabot[bot] <support@github.com> Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com> fix: Resolve race condition in check reducers (openfga#2947) Co-authored-by: Adrian Tam <adrian.tam@okta.com> fix: gateway grpc client tls cert rotation (openfga#2951) Signed-off-by: Shashank Goel <goelshashank13@gmail.com> fix: disable LO pipeline if ff has it set for a store (openfga#2945) chore(deps): bump the dependencies group across 1 directory with 4 updates (openfga#2949) Signed-off-by: dependabot[bot] <support@github.com> Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com> fix: various random fixes (openfga#2942) refactor: check using weighted graph (openfga#2816) Co-authored-by: Yissell Garma <yissell.garma@okta.com> release: update changelog for release `v1.11.6` (openfga#2939) feat: enable pipeline algorithm by default (openfga#2921) add tests to exercise reported bug (openfga#2934) Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com> Co-authored-by: Adrian Tam <adrian.tam@okta.com> refine AGENTS.md to make it more concise with higher value. (openfga#2936) Co-authored-by: Justin Cohen <justincoh@gmail.com> migrate grpc dial context to NewClient (openfga#2714) Co-authored-by: Adrian Tam <adrian.tam@okta.com> Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com> chore(deps): bump filippo.io/edwards25519 from 1.1.0 to 1.1.1 (openfga#2935) Signed-off-by: dependabot[bot] <support@github.com> Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>
Check for parent context cancellation at the beginning of each check reducer (union, intersection, exclusion) to eliminate the race condition between context cancellation and the execution of nested handlers.
Description
What problem is being solved?
The Check call exhibits non-deterministic behavior depending on reducer goroutine scheduling.
Although it typically returns the correct result, it may occasionally fail with a
Request Cancelederror.The issue can be reproduced by running the existing tests with -count=100.
How is it being solved?
Each reducer handler now checks for parent context cancellation. This prevents the race condition between context cancellation and the execution of nested handlers.
Other options:
BenchmarkOpenFGAServer/BenchmarkSQLiteDatastore/BenchmarkCheck/race_between_direct_and_userset.TestRequestContextPropagationwould fail.context.Cancelederrors fromoutcome.errin the reducers. This approach is correct and fast, but requires at least one handler to run to completion.ctx.Err() != nilguard to theconcurrency.NewPoolfunction. This might be a more robust solution, but it would require changes across the entire project.What changes are made to solve it?
ctx.Err() != nilguard to the check reducers (union, intersection, exclusion) to detect parent context cancellation before executing nested handlers.References
Review Checklist
mainSummary by CodeRabbit