MartijnVisser opened a new pull request, #28735:
URL: https://github.com/apache/flink/pull/28735
## What is the purpose of the change
Requesting `/taskmanagers/:taskmanagerid` for a TaskManager that is no
longer registered (for example after a TaskManager is lost while the Web UI's
TaskManager page keeps polling it, or when following a link from a job's
exception history) returned HTTP 500 with an `Unhandled exception` ERROR log on
the JobManager, and the Web UI kept re-requesting it indefinitely.
This returns a proper 404 for a gone TaskManager and stops the futile
polling, showing a clear "TaskManager not available" notice instead of an
endless skeleton.
This supersedes #22542 (thanks @JunRuiLee for the original investigation).
That PR addressed FLINK-31931 by preventing the exception-history page from
linking to a gone TaskManager. Handling the not-found at the destination
instead is more robust: it also covers deep-links and a TaskManager dying while
its page is open, it is not racy (a TaskManager can disappear right after the
link is rendered), and it keeps the link informative for the common case where
the exception's TaskManager is already gone by the time the history is viewed.
## Brief change log
- **[runtime/rest]** `TaskManagerDetailsHandler` strips the failure with
`stripCompletionException` (not `stripExecutionException`), so the
`UnknownTaskExecutorException` to `NOT_FOUND` mapping actually fires; this
matches the sibling TaskManager handlers. Previously a gone TaskManager
produced a 500 plus a repeated `Unhandled exception` ERROR log.
- **[runtime-web]** `loadManager` no longer swallows errors; a shared poll
helper stops polling once the backend reports the TaskManager is gone (404),
while transient errors keep retrying. The TaskManager status view shows a
"TaskManager not available" notice and the metric cards are hidden instead of
rendering blank tables.
## Verifying this change
- New unit test
`TaskManagerDetailsHandlerTest#testUnknownTaskExecutorLeadsToNotFound` asserts
a 404 for an unknown TaskManager id.
- New Vitest specs cover the poll helper (stops on 404, keeps retrying on
transient errors) and the status/metrics components (notice shown / cards
hidden on 404).
- Manually verified on a standalone cluster: killing a TaskManager now
yields a single 404 that stops repeating, a "TaskManager not available" notice,
and no `Unhandled exception` ERROR log on the JobManager, versus the previous
endless 500s.
## Does this pull request potentially affect one of the following parts:
- Dependencies (does it add or upgrade a dependency): no
- The public API, i.e., is any changed class annotated with
`@Public(Evolving)`: no
- The serializers: no
- The runtime per-record code paths (performance sensitive): no
- Anything that affects deployment or recovery: no
- The S3 file system connector: no
## Documentation
- Does this pull request introduce a new feature? no
- If yes, how is the feature documented? not applicable
## AI Usage Disclosure
- [x] Claude Code
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
