rls: delegate pick to child policy as long as it is not in TransientFailure #5656
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
easwars
force-pushed
the
e2e_with_ring_hash
branch
from
3907d39 to
da5a854
Compare
dfawley
reviewed
Sep 15, 2022
balancer/rls/picker.go
Outdated
| // to which this RPC can be routed to. If there is no child policy in READY | ||
| // state, we delegate to the first child policy arbitrarily. | ||
| // to which this RPC can be routed to. If all child policies are in | ||
| // TRANSIENT_FAILURE, we delegate to the first child policy arbitrarily. |
Member
There was a problem hiding this comment.
Should we delegate to the last child policy to match C++? (Change itself should be ~trivial; e.g. check if loop iterator (currently _) is len(dcEntry.childPolicyWrappers)-1, in which case don't continue even if TRANSIENT_FAILURE.)
Contributor
Author
There was a problem hiding this comment.
I thought about it when making the change, but was lazy to make it and thought you wouldn't ask for it ... lol
dfawley
approved these changes
Sep 15, 2022
Sign up for free
to subscribe to this conversation on GitHub.
Already have an account?
Sign in.
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Outdated suggestions cannot be applied.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.
Recently, we implemented a no-op
ExitIdlemethod for the ringhash LB policy in #5614.When RLS is used in conjunction with ringhash though, this causes the channel to never come out of
IDLE, as ringhash starts inIDLEand any requests to exit idle from its parent results in a no-op.The pick behavior in RLS though seems to be the root cause. Instead of delegating the picks to child policies which are not in
TransientFailure, RLS was asking them to exit idle if they were inIDLE. This is in contrast to how things are implemented in c-core, which simply delegates the picks to the child policies as long as they are not inTransientFailure, and for a policy likeringhash, the pick ensures that it exits idle.RELEASE NOTES:
IDLE