Skip to content

Flaky test: parallel/test-tls-ticket-cluster #29341

Closed as not planned
Closed as not planned
Flaky test: parallel/test-tls-ticket-cluster#29341
Labels
child_processIssues and PRs related to the child_process subsystem.flaky-testIssues and PRs involving tests that fail intermittently in CI.macosIssues and PRs related to the macOS platform.staleIssues and PRs marked stale due to inactivity and scheduled for automatic closure.tlsIssues and PRs related to the tls subsystem.

Description

@mscdex
mscdex
opened on Aug 27, 2019
  • Version: master
  • Platform: test-macstadium-macos10.11-x64-1
  • Subsystem: test

Console output

04:34:35 not ok 2094 parallel/test-tls-ticket-cluster
04:34:35   ---
04:34:35   duration_ms: 0.579
04:34:35   severity: fail
04:34:35   exitcode: 1
04:34:35   stack: |-
04:34:35     [master] got "listening"
04:34:35     [master] connecting 60380 session? false
04:34:35     [worker] connection reused? false
04:34:35     [master] got "not-reused"
04:34:35     [master] connecting 60380 session? true
04:34:35     [worker] connection reused? true
04:34:35     [master] got "reused"
04:34:35     [master] connecting 60380 session? true
04:34:35     [worker] connection reused? true
04:34:35     [master] got "reused"
04:34:35     [master] connecting 60380 session? true
04:34:35     [worker] connection reused? true
04:34:35     [master] got "reused"
04:34:35     [master] connecting 60380 session? true
04:34:35     [worker] connection reused? true
04:34:35     [master] got "reused"
04:34:35     [master] connecting 60380 session? true
04:34:35     [worker] connection reused? true
04:34:35     [master] got "reused"
04:34:35     [master] connecting 60380 session? true
04:34:35     [worker] connection reused? true
04:34:35     [master] got "reused"
04:34:35     [master] connecting 60380 session? true
04:34:35     [worker] connection reused? true
04:34:35     [master] got "reused"
04:34:35     [master] connecting 60380 session? true
04:34:35     [worker] connection reused? true
04:34:35     [master] got "reused"
04:34:35     [master] connecting 60380 session? true
04:34:35     [worker] connection reused? true
04:34:35     [master] got "reused"
04:34:35     [master] connecting 60380 session? true
04:34:35     [worker] connection reused? true
04:34:35     [master] got "reused"
04:34:35     [master] connecting 60380 session? true
04:34:35     [master] got "listening"
04:34:35     [worker] connection reused? true
04:34:35     [master] got "reused"
04:34:35     [master] connecting 60380 session? true
04:34:35     [worker] connection reused? true
04:34:35     [master] got "reused"
04:34:35     [master] connecting 60380 session? true
04:34:35     [worker] connection reused? true
04:34:35     [master] got "reused"
04:34:35     [master] connecting 60380 session? true
04:34:35     [worker] connection reused? true
04:34:35     [master] got "reused"
04:34:35     [master] connecting 60380 session? true
04:34:35     [worker] connection reused? true
04:34:35     [master] got "reused"
04:34:35     [worker] got "die"
04:34:35     [worker] server close
04:34:35     [worker] exit
04:34:35     [worker] got "die"
04:34:35     [worker] server close
04:34:35     [worker] exit
04:34:35     [worker] got "die"
04:34:35     [worker] server close
04:34:35     [worker] exit
04:34:35     [worker] got "die"
04:34:35     [worker] server close
04:34:35     [worker] exit
04:34:35     events.js:186
04:34:35           throw er; // Unhandled 'error' event
04:34:35           ^
04:34:35     
04:34:35     Error: write EPIPE
04:34:35         at ChildProcess.target._send (internal/child_process.js:813:20)
04:34:35         at ChildProcess.target.send (internal/child_process.js:683:19)
04:34:35         at sendHelper (internal/cluster/utils.js:22:15)
04:34:35         at send (internal/cluster/master.js:351:10)
04:34:35         at internal/cluster/master.js:317:5
04:34:35         at done (internal/cluster/round_robin_handle.js:46:7)
04:34:35         at RoundRobinHandle.add (internal/cluster/round_robin_handle.js:55:12)
04:34:35         at queryServer (internal/cluster/master.js:311:10)
04:34:35         at Worker.onmessage (internal/cluster/master.js:246:5)
04:34:35         at ChildProcess.onInternalMessage (internal/cluster/utils.js:43:8)
04:34:35     Emitted 'error' event on Worker instance at:
04:34:35         at ChildProcess.<anonymous> (internal/cluster/worker.js:27:12)
04:34:35         at ChildProcess.emit (events.js:209:13)
04:34:35         at internal/child_process.js:817:39
04:34:35         at processTicksAndRejections (internal/process/task_queues.js:75:11) {
04:34:35       errno: -32,
04:34:35       code: 'EPIPE',
04:34:35       syscall: 'write'
04:34:35     }

Activity

  1. mscdex
    added
    flaky-testIssues and PRs involving tests that fail intermittently in CI.
    tlsIssues and PRs related to the tls subsystem.
    on Aug 27, 2019

  2. mscdex commented on Aug 28, 2019

    @mscdex
    mscdex
    on Aug 28, 2019
    ContributorAuthor

    /cc @nodejs/cluster ?

  3. Trott commented on Aug 28, 2019

    @Trott
    Trott
    on Aug 28, 2019
    Member

    Also on FreeBSD:

    https://ci.nodejs.org/job/node-test-commit-freebsd/28120/nodes=freebsd11-x64/console

    00:15:03 not ok 1983 parallel/test-tls-ticket-cluster
00:15:03   ---
00:15:03   duration_ms: 0.865
00:15:03   severity: fail
00:15:03   exitcode: 1
00:15:03   stack: |-
00:15:03     [master] got "listening"
00:15:03     [master] connecting 28298 session? false
00:15:03     [worker] connection reused? false
00:15:03     [master] got "not-reused"
00:15:03     [master] connecting 28298 session? true
00:15:03     [worker] connection reused? true
00:15:03     [master] got "reused"
00:15:03     [master] connecting 28298 session? true
00:15:03     [worker] connection reused? true
00:15:03     [master] got "reused"
00:15:03     [master] connecting 28298 session? true
00:15:03     [worker] connection reused? true
00:15:03     [master] got "reused"
00:15:03     [master] connecting 28298 session? true
00:15:03     [worker] connection reused? true
00:15:03     [master] got "reused"
00:15:03     [master] connecting 28298 session? true
00:15:03     [worker] connection reused? true
00:15:03     [master] got "reused"
00:15:03     [master] connecting 28298 session? true
00:15:03     [worker] connection reused? true
00:15:03     [master] got "reused"
00:15:03     [master] connecting 28298 session? true
00:15:03     [worker] connection reused? true
00:15:03     [master] got "reused"
00:15:03     [master] connecting 28298 session? true
00:15:03     [worker] connection reused? true
00:15:03     [master] got "reused"
00:15:03     [master] connecting 28298 session? true
00:15:03     [worker] connection reused? true
00:15:03     [master] got "reused"
00:15:03     [master] connecting 28298 session? true
00:15:03     [worker] connection reused? true
00:15:03     [master] got "reused"
00:15:03     [master] connecting 28298 session? true
00:15:03     [worker] connection reused? true
00:15:03     [master] got "reused"
00:15:03     [master] connecting 28298 session? true
00:15:03     [worker] connection reused? true
00:15:03     [master] got "reused"
00:15:03     [master] connecting 28298 session? true
00:15:03     [worker] connection reused? true
00:15:03     [master] got "reused"
00:15:03     [master] connecting 28298 session? true
00:15:03     [worker] connection reused? true
00:15:03     [master] got "reused"
00:15:03     [master] connecting 28298 session? true
00:15:03     [worker] connection reused? true
00:15:03     [master] got "reused"
00:15:03     [worker] got "die"
00:15:03     [worker] server close
00:15:03     [worker] exit
00:15:03     [master] worker died
00:15:03     [worker] got "die"
00:15:03     [worker] server close
00:15:03     [worker] exit
00:15:03     [worker] got "die"
00:15:03     [worker] server close
00:15:03     [worker] exit
00:15:03     [master] worker died
00:15:03     [worker] got "die"
00:15:03     [worker] server close
00:15:03     [worker] exit
00:15:03     events.js:186
00:15:03           throw er; // Unhandled 'error' event
00:15:03           ^
00:15:03     
00:15:03     Error: write EPIPE
00:15:03         at ChildProcess.target._send (internal/child_process.js:813:20)
00:15:03         at ChildProcess.target.send (internal/child_process.js:683:19)
00:15:03         at sendHelper (internal/cluster/utils.js:22:15)
00:15:03         at send (internal/cluster/master.js:351:10)
00:15:03         at internal/cluster/master.js:317:5
00:15:03         at Server.done (internal/cluster/round_robin_handle.js:46:7)
00:15:03         at Object.onceWrapper (events.js:298:28)
00:15:03         at Server.emit (events.js:214:15)
00:15:03         at emitListeningNT (net.js:1332:10)
00:15:03         at processTicksAndRejections (internal/process/task_queues.js:79:21)
00:15:03     Emitted 'error' event on Worker instance at:
00:15:03         at ChildProcess.<anonymous> (internal/cluster/worker.js:27:12)
00:15:03         at ChildProcess.emit (events.js:209:13)
00:15:03         at internal/child_process.js:817:39
00:15:03         at processTicksAndRejections (internal/process/task_queues.js:75:11) {
00:15:03       errno: -32,
00:15:03       code: 'EPIPE',
00:15:03       syscall: 'write'
00:15:03     }
00:15:03   ...

  4. Trott commented on Aug 28, 2019

    @Trott
    Trott
    on Aug 28, 2019
    Member

    I'm so far unable to replicate this locally, but EPIPE can happen if you try to write to a closed stream, and there have been some changes to streams lately... @nodejs/streams? @ronag?

  5. lpinca commented on Aug 31, 2019

    @lpinca
    lpinca
    on Aug 31, 2019
    Member

    I can reproduce with

    $ ./tools/test.py -J --repeat=500 test/parallel/test-tls-ticket-cluster.js

    and bisecting points to 4367732 as first bad commit.

    @nodejs/libuv

  6. lpinca
    added
    child_processIssues and PRs related to the child_process subsystem.
    macosIssues and PRs related to the macOS platform.
    on Aug 31, 2019

  7. bnoordhuis commented on Aug 31, 2019

    @bnoordhuis
    bnoordhuis
    on Aug 31, 2019
    Member

    That probably means libuv/libuv@ee24ce9 is responsible (cc @addaleax) but that also means it's probably exposing a latent bug in Node.js (edit: or the test), not introducing a new one.

    The sole uv_try_write() caller is below and that err == UV_EAGAIN check looks highly suspect:

    node/src/stream_wrap.cc

    Lines 350 to 354 in 98b7185

    err = uv_try_write(stream(), vbufs, vcount);
    if (err == UV_ENOSYS || err == UV_EAGAIN)
    return 0;
    if (err < 0)
    return err;

  8. addaleax commented on Aug 31, 2019

    @addaleax
    addaleax
    on Aug 31, 2019
    Member

    It looks more like issues with the cluster/child process code to me… currently, synchronous write errors from .send() are reported (but there is an ancient undocumented option to turn that off, swallowErrors), while asynchronous write errors are always discarded silently.

    It generally also looks like there’s very little error handling in the cluster module in general, and in particular, there’s this bit that’s also in the stack trace:

    node/lib/internal/cluster/round_robin_handle.js

    Lines 45 to 48 in 98b7185

    // TODO(bnoordhuis) Check err.
    send(null, { sockname: out }, null);
    } else {
    send(null, null, null); // UNIX socket.

    Judging from the commit message of 0da4e0e (which added swallowErrors), I’d think it make sense to add that flag for internal cluster messages, too.

    /cc @cjihrig

    The sole uv_try_write() caller is below and that err == UV_EAGAIN check looks highly suspect:

    I think UV_ENOSYS and UV_EAGAIN are precisely the return values that indicate that it makes sense to use uv_write() after a failed uv_try_write(), correct?

  9. bnoordhuis commented on Sep 1, 2019

    @bnoordhuis
    bnoordhuis
    on Sep 1, 2019
    Member

    I think UV_ENOSYS and UV_EAGAIN are precisely the return values that indicate that it makes sense to use uv_write() after a failed uv_try_write(), correct?

    Yes, but before libuv/libuv@ee24ce9 that path was always taken. The if (err < 0) below was dead code but now it's not.

  10. MylesBorins commented on Apr 1, 2020

    @MylesBorins
    MylesBorins
    on Apr 1, 2020
    Contributor

    This failure is showing up on v10.x now, specifically on centos7. Did this get figured out on master? Should we mark it as flaky?

  11. Trott commented on Apr 4, 2020

    @Trott
    Trott
    on Apr 4, 2020
    Member

    This failure is showing up on v10.x now, specifically on centos7. Did this get figured out on master? Should we mark it as flaky?

    Using ncu-ci walk, I can only find one recent example, and it's a node-daily-v10.x-staging daily job.

  12. puzpuzpuz commented on Apr 23, 2020

    @puzpuzpuz
    puzpuzpuz
    on Apr 23, 2020
    Member

    Saw this one failing on OS X today: https://ci.nodejs.org/job/node-test-commit-osx/33696/nodes=osx1015/testReport/junit/(root)/test/parallel_test_tls_ticket_cluster/

  13. github-actions commented on Jun 27, 2026

    @github-actions
    github-actionsbot
    on Jun 27, 2026 – with GitHub Actions
    Contributor

    This issue has been marked as stale due to 210 days of inactivity.
    It will be automatically closed in 30 days if no further activity occurs. If this is still relevant, please leave a comment or update it to keep it open.

  14. github-actions
    added
    staleIssues and PRs marked stale due to inactivity and scheduled for automatic closure.
    on Jun 27, 2026

  15. github-actions commented on Jul 28, 2026

    @github-actions
    github-actionsbot
    on Jul 28, 2026 – with GitHub Actions
    Contributor

    This issue has been automatically closed after 30 days of inactivity following its stale status (no activity for a total of 120 days).
    If this is still relevant, feel free to reopen it or leave a comment with additional details so we can continue the discussion.

  16. github-actions
    closed this as not plannedon Jul 28, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    child_processIssues and PRs related to the child_process subsystem.flaky-testIssues and PRs involving tests that fail intermittently in CI.macosIssues and PRs related to the macOS platform.staleIssues and PRs marked stale due to inactivity and scheduled for automatic closure.tlsIssues and PRs related to the tls subsystem.

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions