-
[OPIK-6982] Fix ClosingTraceThreadSubscriber NPE on null workspace timeout (#7147)
frostbyte_neo 发布于
2026-06-18 09:27:43 +00:00 - [OPIK-6982] Fix ClosingTraceThreadSubscriber NPE on null workspace timeout
The closing trace-thread subscriber failed with
"NullPointerException: The mapper [...] returned a null value." for any
workspace that has a configuration record where timeoutToMarkThreadAsInactive
is null (e.g. only truncationOnTables is set).getWorkspaceTimeout used
.map(WorkspaceConfiguration::timeoutToMarkThreadAsInactive)
followed by.switchIfEmpty(default). switchIfEmpty only fires on empty
completion, not when the config is present but the timeout field is null —
so map() emitted null, which Reactor rejects, and that workspace's threads
never closed.Switch to
.mapNotNull(...).defaultIfEmpty(default)so a null timeout drops
to empty and falls back to the default, covering both "no config" and
"config with null timeout".Adds a regression test reproducing the null-timeout workspace config.
Co-Authored-By: Claude Opus 4.8 (1M context) noreply@anthropic.com
- [OPIK-6982] Bound regression test await with atMost so it fails fast
Address review feedback (Baz): the Awaitility chain in the new test was
unbounded and could hang indefinitely if the closing job never ran. Add
.atMost(30, TimeUnit.SECONDS), matching the backend test convention.Co-Authored-By: Claude Opus 4.8 (1M context) noreply@anthropic.com
- [OPIK-6982] Derive closing-job test waits from configured timeout
Address review nit (JetoPistola): the closure-wait delays were hardcoded
literals coupled to traceThreadConfig.timeoutToMarkThreadAsInactive in
config-test.yml, so all three tests would silently drift if that default
changed.Inject the @Config-bound TraceThreadConfig and derive the waits:
- default-timeout cases wait defaultThreadTimeout + 500ms buffer
- custom-timeout case derives its waits from the local customTimeout
Also bound the remaining await() chains with .atMost(30s) so they fail fast.
Co-Authored-By: Claude Opus 4.8 (1M context) noreply@anthropic.com
- [OPIK-6982] Address review nits in closing-job test
- Use UUID v7 for thread IDs (matches production) via a TimeBasedEpochGenerator.
- Replace Mono.delay(...).block() sleeps with TestUtils.waitForMillis(...).
- Bound every await() with an aggressive .atMost(AWAIT_TIMEOUT_SECONDS=10s)
so tests fail fast (Awaitility already defaults to 10s; making it explicit
and uniform, including the previously unbounded reopen-test awaits).
Co-Authored-By: Claude Opus 4.8 (1M context) noreply@anthropic.com
Co-authored-by: Claude Opus 4.8 (1M context) noreply@anthropic.com
下载附件