Skip to content

[Inference API] Fix test for pre-adaptive rate limiting case #124709

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 1 commit into from
Mar 18, 2025
Merged

[Inference API] Fix test for pre-adaptive rate limiting case #124709

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
25 changes: 18 additions & 7 deletions .../test/java/org/elasticsearch/xpack/core/inference/action/InferenceActionRequestTests.java
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -318,8 +318,10 @@ protected InferenceAction.Request mutateInstance(InferenceAction.Request instanc

@Override
protected InferenceAction.Request mutateInstanceForVersion(InferenceAction.Request instance, TransportVersion version) {
InferenceAction.Request mutated;

if (version.before(TransportVersions.V_8_12_0)) {
return new InferenceAction.Request(
mutated = new InferenceAction.Request(
instance.getTaskType(),
instance.getInferenceEntityId(),
null,
Expand All @@ -330,7 +332,7 @@ protected InferenceAction.Request mutateInstanceForVersion(InferenceAction.Reque
false
);
} else if (version.before(TransportVersions.V_8_13_0)) {
return new InferenceAction.Request(
mutated = new InferenceAction.Request(
instance.getTaskType(),
instance.getInferenceEntityId(),
null,
Expand All @@ -344,7 +346,7 @@ protected InferenceAction.Request mutateInstanceForVersion(InferenceAction.Reque
&& (instance.getInputType() == InputType.UNSPECIFIED
|| instance.getInputType() == InputType.CLASSIFICATION
|| instance.getInputType() == InputType.CLUSTERING)) {
return new InferenceAction.Request(
mutated = new InferenceAction.Request(
instance.getTaskType(),
instance.getInferenceEntityId(),
null,
Expand All @@ -356,7 +358,7 @@ protected InferenceAction.Request mutateInstanceForVersion(InferenceAction.Reque
);
} else if (version.before(TransportVersions.V_8_13_0)
&& (instance.getInputType() == InputType.CLUSTERING || instance.getInputType() == InputType.CLASSIFICATION)) {
return new InferenceAction.Request(
mutated = new InferenceAction.Request(
instance.getTaskType(),
instance.getInferenceEntityId(),
null,
Expand All @@ -367,7 +369,7 @@ protected InferenceAction.Request mutateInstanceForVersion(InferenceAction.Reque
false
);
} else if (version.before(TransportVersions.V_8_14_0)) {
return new InferenceAction.Request(
mutated = new InferenceAction.Request(
instance.getTaskType(),
instance.getInferenceEntityId(),
null,
Expand All @@ -379,7 +381,7 @@ protected InferenceAction.Request mutateInstanceForVersion(InferenceAction.Reque
);
} else if (version.before(TransportVersions.INFERENCE_CONTEXT)
&& version.isPatchFrom(TransportVersions.INFERENCE_CONTEXT_8_X) == false) {
return new InferenceAction.Request(
mutated = new InferenceAction.Request(
instance.getTaskType(),
instance.getInferenceEntityId(),
instance.getQuery(),
Expand All @@ -390,9 +392,18 @@ protected InferenceAction.Request mutateInstanceForVersion(InferenceAction.Reque
false,
InferenceContext.EMPTY_INSTANCE
);
} else {
mutated = instance;
}

return instance;
// We always assume that a request has been rerouted, if it came from a node without adaptive rate limiting
if (version.before(TransportVersions.INFERENCE_REQUEST_ADAPTIVE_RATE_LIMITING)) {
mutated.setHasBeenRerouted(true);
} else {
mutated.setHasBeenRerouted(instance.hasBeenRerouted());
}

return mutated;
}

public void testWriteTo_WhenVersionIsOnAfterUnspecifiedAdded() throws IOException {
Expand Down