Deterministic Subsetting Lb Policy #10470
Conversation
joybestourous
changed the title
| * https://github.com/grpc/proposal/blob/master/A68-deterministic-subsetting-lb-policy.md | ||
| */ | ||
| @Internal | ||
| public final class DeterministicSubsettingLoadBalancer extends LoadBalancer { |
There was a problem hiding this comment.
Remove the public (and thus @Internal). Only the Provider needs to be public.
| (DeterministicSubsettingLoadBalancerConfig) | ||
| resolvedAddresses.getLoadBalancingPolicyConfig(); | ||
|
|
||
| switchLb.switchTo(config.childPolicy.getProvider()); |
There was a problem hiding this comment.
Since you made this change, I've cleaned this up and switchTo() is now deprecated. Instead of switchTo, you'll just use setLoadBalancingPolicyConfig() and pass a GracefulSwitchLB configuration object. This should clean up the parsing in the logic in the provider, as now GracefulSwitchLB.parseLoadBalancingPolicyConfig() does much of the work instead of random internal APIs.
2c49cc4 might be a reasonable example of the change.
|
|
||
| @VisibleForTesting | ||
| static class AddressComparator implements Comparator<SocketAddress> { | ||
| // For consistency with the golang instrumentation, this assumes toString is overridden such |
There was a problem hiding this comment.
This isn't consistent with golang. I know this will change to hashing, but there will still be some similar issues. Maybe we'll be forced to use address.hashCode(), or maybe it doesn't matter because we just have to be consistent within a single implementation. See f866c80 where I mention some issues with the string representation.
|
@joybestourous Are you still interested in getting this PR merged? |
A POC for grpc/proposal#383