hyperledger / fabric-x-committer

91%

main: 92%

LAST BUILD ON BRANCH fix-flaky

branch: fix-flaky

CHANGE BRANCH

x

Reset

• fix-flaky
• apt-ci-fix
• backoff-v5
• bccsp-race
• bench-report-metrics
• bft-delivery
• bob-rules
• buffer-race
• build-unify
• channel-race
• ci-gotestsum
• cli-refactor
• client-test-race
• client-tls-bug
• commit-config-immediately
• config-docs-yaml
• connection-test-dep
• coveralls
• cryptogen-recreate-support
• default-test-node-startup-setup
• deliver-check-block-order
• deliver-seek-tls-cert
• deliver-seek-tls-cert-fix
• dependabot/go_modules/go_modules-9c5197dcb8
• dependencygraph-race
• deployment-strategy
• doc-loadgen-artifacts
• dynamic-root-ca
• enable-pprof
• expose-ledger-api
• extract-prometheus-startup-and-grpc-server-start
• fabric-lib-go-v1.1.3
• fabric-logging
• fix-connection-reset
• fix-container-image
• fix-container-test-macos
• fix-db-test
• fix-falky-svmupdate-test
• fix-flaky-conn-metrics
• fix-flaky-loadgen-coord
• fix-master-tserver-rf
• fix-port-container-test
• fix-release-path
• fix-sign
• fix-test-node-with-insecure
• fix-waiting-tx-count
• flaky-test-orderer
• getpeersmspdirs-no-hardcode
• go-fix
• grpc-logger-fix
• grpcservice-refactor
• healthz-check
• ledger-service-wait-behaviour
• lifecycle-policy
• limit-block-apis
• limit-inputs-block-apis
• load-config-block-certs
• loadgen-manual-msp
• loadgen-remove-ctx
• main
• make-ci-local
• make-fixes
• makefile-test-flags
• metrics-tls
• metrics-url-race
• migrate-to-postgres-official-image
• minimize-logs
• mkdocs
• mock-orderer-features
• mock-orderer-stream-holder
• mock-verifier-race
• move-retry-package
• namespace-policy-docs
• optimize-map-block
• orderer-config-race
• pgx5
• postgres-renaming
• qs-pen-test-fix
• race-detection
• rate-limit-sidecar
• refactor-channel
• refactor-cmd
• rename-util-files
• review-ai-prompt
• rm-reployment
• rm-unused
• root-ca-workaround-fix
• sidecar-block-store-code-fix
• sidecar-limit-active-txids-notifier
• sidecar-race
• sidecar-remove-chan-id
• split-tablets
• sustain-config
• sync-block
• template-tls-per-endpoint
• test-container-port-binding
• test-coverage-connection
• test-coverage-improve
• test-coverage-latency
• test-coverage-tests
• test-limiter-max-wait
• test-minor-fixes
• test-query-status
• test-tls-flatten
• testcrypto-race
• testcrypto-refactor
• tls-docs-update
• update-common-dep
• update-to-latest-common
• update-to-yugabytedb-latest-lts
• upgrade-pg
• vc-race
• verifier-test-remove-gomega
• worktree-notification-docs

Build # 23794732970

Build Type

Pull #501

github

Committed by cendhu

Commit Message

[test] fix flaky TestSignatureVerifierManagerPolicyUpdateAndRecover

Root cause: requireAllUpdate submits many batches (3 per 10ms tick),
filling the outputValidatedTxs channel (capacity 10). When the channel
is full, signVerifier goroutines block on Write() instead of being in
Recv(). Both goroutines in the errgroup deadlock: receiveStatus blocks
on Write (channel full), sendTransactions blocks on Read (no batches),
and neither can cancel gCtx. The drain goroutine was started after
requireAllUpdate, so the channel was already full by the time it ran.
On busy CI machines, the drain goroutine's scheduling delay combined
with buffered responses meant signVerifier[0] could remain stuck on
Write for the entire 30-second EventuallyIntMetric timeout, never
reaching Recv() to detect the broken stream from server.Stop().

Fix: move the drain goroutine before requireAllUpdate so the output
channel never fills up. This ensures signVerifiers are always in their
idle state (blocked on Recv) when the server stops, allowing immediate
stream break detection. Use channel.Ready to guarantee the drain
goroutine is running before proceeding.

Signed-off-by: Senthilnathan <cendhu@gmail.com>

Pull Request Pull Request #501: [test] fix flaky TestSignatureVerifierManagerPolicyUpdateAndRecover

Coverage Stats

9936 of 10872 relevant lines covered (91.39%)

42282.43 hits per line