`groupBy`'s default behavior is to filter elements, that would go to a substream
that is already closed. This allows `groupBy` to recreate those already closed streams,
if wanted, and run them as if they would've never run before in the first place.
Overload instead of default parameter
Make set of closed keys unmodifiable if not needed
Adjust documentation, adding a warning on memory consumption.
Add MiMa exclude.
* Documented wrapping PersistentBehavior in typed actors persistence documentation #24679
* Removed "Current Limitations" section
* Added description on wrapping PersistentBehavior
* Added Scala and Java example for the pattern
* updated persistence docs for the hinted changes
* added missing "the" as requested in the hint
* RestartWithBackOff delay cancel to wait for failure
For wrapping a user flow with FlowRestart.onFlowWithFailures when the user
flow it fails signals a cancel upstream and a failure downstream.
These are intercepted by a SubSource/SubSink. In the case
the SubSource receives the cancel before the SubSink receives
the real upstream is wrongly canceled leading to an error
when the SubSink restarts the flow.
This commit introduces a delay for the cancel so that the failure
is more likely to win.
Would be far better to propagate a reason for cancel so this could
be deterministic. See https://github.com/akka/akka/pull/23909
Refs #24528#24726
* Allow reuse of journal loading logic
Probably can be generialised even more. At present i've tested this with
the akka persistence updater extension.
* Make the implicit explicit
* Added mima filter for removing internal class
* Review feedback
* travis config from akka-http
* amend CONTRIBUTING to explain our tests
* amend wording in CONTRIBUTING
* new whitesource key
* remove instead of comment deployment things
* Update CONTRIBUTING.md
* exclude new module in 2.5 from appearing in 2.4 previous mima artifacts
* fix mima on 2.12
The test failed timing out waiting for a cluster to remove a member. It
was nearly done. This change uses the common configuration used for
cluster tests to speed up membership changes rather than increasing the
timeout.
Each build is now over 40mb logs.
A lot of DEBUG logging was left on for test failures that have been
fixed. Added an issue # for ones that are still valid or if if it on
as the test verifies debug
`Sink.contramap(f)` simply applies a function and then materializes using the original sink. It would seem very odd if backpressure would cause the Flow to cancel...
