grpclb: fallback timer only when not already using fallback backends. #8646
Conversation
This file contains 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
…java into already_in_fallback
ejona86
approved these changes
Nov 2, 2021
There was a problem hiding this comment.
That was quite difficult to find. Good thinking for considering the case that led us to finding this!
For a service in steady-state, this should be somewhat hard to hit as it would require a partial DNS failure (for SRV lookup to fail) and then for the grpclb response to be slow (either network or slow server). That will happen, but isn't all that likely.
But if you consider the case of a service enabling gRPC-LB, all of a sudden it only requires the grpclb response to be slow. That is quite scary. We should definitely backport this, potentially to all affected versions (although we'll have to discuss what versions we'll do a release for).
temawi
added
the
TODO:backport
temawi
added a commit
to temawi/grpc-java
that referenced
this pull request
Nov 2, 2021
…grpc#8646) Addresses a problem where we initially only resolve addresses to the backends, but not the load balancer and then later resolve addresses to both. In this situation the fallback timer was started during the second instance even if it resulted in the timer later failing as we were already using fallback backends. This change assures that a fallback time is only ever started if we are not already using the fallback backends. This is a follow-up fix to grpc#8253.
temawi
added a commit
that referenced
this pull request
Nov 2, 2021
…#8646) (#8648) Addresses a problem where we initially only resolve addresses to the backends, but not the load balancer and then later resolve addresses to both. In this situation the fallback timer was started during the second instance even if it resulted in the timer later failing as we were already using fallback backends. This change assures that a fallback time is only ever started if we are not already using the fallback backends. This is a follow-up fix to #8253.
temawi
added a commit
to temawi/grpc-java
that referenced
this pull request
Nov 3, 2021
…grpc#8646) Addresses a problem where we initially only resolve addresses to the backends, but not the load balancer and then later resolve addresses to both. In this situation the fallback timer was started during the second instance even if it resulted in the timer later failing as we were already using fallback backends. This change assures that a fallback time is only ever started if we are not already using the fallback backends. This is a follow-up fix to grpc#8253.
temawi
added a commit
to temawi/grpc-java
that referenced
this pull request
Nov 3, 2021
…grpc#8646) Addresses a problem where we initially only resolve addresses to the backends, but not the load balancer and then later resolve addresses to both. In this situation the fallback timer was started during the second instance even if it resulted in the timer later failing as we were already using fallback backends. This change assures that a fallback time is only ever started if we are not already using the fallback backends. This is a follow-up fix to grpc#8253.
temawi
added a commit
to temawi/grpc-java
that referenced
this pull request
Nov 3, 2021
…grpc#8646) Addresses a problem where we initially only resolve addresses to the backends, but not the load balancer and then later resolve addresses to both. In this situation the fallback timer was started during the second instance even if it resulted in the timer later failing as we were already using fallback backends. This change assures that a fallback time is only ever started if we are not already using the fallback backends. This is a follow-up fix to grpc#8253.
temawi
added a commit
to temawi/grpc-java
that referenced
this pull request
Nov 3, 2021
…grpc#8646) Addresses a problem where we initially only resolve addresses to the backends, but not the load balancer and then later resolve addresses to both. In this situation the fallback timer was started during the second instance even if it resulted in the timer later failing as we were already using fallback backends. This change assures that a fallback time is only ever started if we are not already using the fallback backends. This is a follow-up fix to grpc#8253.
temawi
added a commit
that referenced
this pull request
Nov 3, 2021
…#8646) (#8651) Addresses a problem where we initially only resolve addresses to the backends, but not the load balancer and then later resolve addresses to both. In this situation the fallback timer was started during the second instance even if it resulted in the timer later failing as we were already using fallback backends. This change assures that a fallback time is only ever started if we are not already using the fallback backends. This is a follow-up fix to #8253.
temawi
added a commit
that referenced
this pull request
Nov 3, 2021
…#8646) (#8653) Addresses a problem where we initially only resolve addresses to the backends, but not the load balancer and then later resolve addresses to both. In this situation the fallback timer was started during the second instance even if it resulted in the timer later failing as we were already using fallback backends. This change assures that a fallback time is only ever started if we are not already using the fallback backends. This is a follow-up fix to #8253.
temawi
added a commit
that referenced
this pull request
Nov 3, 2021
…#8646) (#8652) Addresses a problem where we initially only resolve addresses to the backends, but not the load balancer and then later resolve addresses to both. In this situation the fallback timer was started during the second instance even if it resulted in the timer later failing as we were already using fallback backends. This change assures that a fallback time is only ever started if we are not already using the fallback backends. This is a follow-up fix to #8253.
temawi
added a commit
that referenced
this pull request
Nov 3, 2021
…#8646) (#8654) Addresses a problem where we initially only resolve addresses to the backends, but not the load balancer and then later resolve addresses to both. In this situation the fallback timer was started during the second instance even if it resulted in the timer later failing as we were already using fallback backends. This change assures that a fallback time is only ever started if we are not already using the fallback backends. This is a follow-up fix to #8253.
beatrausch
pushed a commit
to beatrausch/grpc-java
that referenced
this pull request
Nov 4, 2021
…grpc#8646) Addresses a problem where we initially only resolve addresses to the backends, but not the load balancer and then later resolve addresses to both. In this situation the fallback timer was started during the second instance even if it resulted in the timer later failing as we were already using fallback backends. This change assures that a fallback time is only ever started if we are not already using the fallback backends. This is a follow-up fix to grpc#8253.
ejona86
removed
the
TODO:backport
Sign up for free
to subscribe to this conversation on GitHub.
Already have an account?
Sign in.
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.
Addresses a problem where we initially only resolve addresses to the backends, but not the load balancer and then later resolve addresses to both. In this situation the fallback timer was started during the second instance even if it resulted in the timer later failing as we were already using fallback backends.
This change assures that a fallback time is only ever started if we are not already using the fallback backends.
This is a follow-up fix to #8253.