[FLINK-5040] [jobmanager] Set correct input channel types with eager scheduling #2783
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
uce
force-pushed
the
eager_deployment
branch
from
9d186d9 to
96b2ee2
Compare
The reverted commit did not really fix anything, but hid the problem by brute force, sending many more schedule or update consumers messages.
The back offs were hard coded before, which would have made it impossible to react to any potential problems with them.
uce
force-pushed
the
eager_deployment
branch
from
96b2ee2 to
ca797d9
Compare
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
When we do eager deployment all intermediate stream/partition locations are already known when scheduling an intermediate stream/partition consumer. Nonetheless we saw tasks with "unknown input channels" that were updated lazily during runtime. This was caused by a wrong producer execution state check requiring the producers to be in RUNNING or DEPLOYING state when creating consumer input channels. This is changed in the 2nd commit.
The 1st commit revert a bogus fix as part of FLINK-3232. With that "fix" we actually did not fix anything correctly and instead doubled the number of schedule or update consumer messages we sent.
Furthermore (3rd commit) we change the initial and max partition request back off to 100ms and 10secs respectively. Those numbers were hard coded before. As a safety net for very slow deployments, the values can be changed via the config. No user should need to change this config value in practice.