test(coordinator): rewire e2e onto shared inference-gateway and add 3-EPP topology - #2253
Open
revit13 wants to merge 6 commits into
Open
test(coordinator): rewire e2e onto shared inference-gateway and add 3-EPP topology#2253revit13 wants to merge 6 commits into
revit13 wants to merge 6 commits into
Conversation
Signed-off-by: Revital Sur <eres@il.ibm.com>
Signed-off-by: Revital Sur <eres@il.ibm.com>
Signed-off-by: Revital Sur <eres@il.ibm.com>
Signed-off-by: Revital Sur <eres@il.ibm.com>
Signed-off-by: Revital Sur <eres@il.ibm.com>
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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
What type of PR is this?
/kind cleanup
/kind test
What this PR does / why we need it:
Reworks the coordinator e2e suite to exercise the disaggregation topologies the deployment guide (llm-d/llm-d#2119) actually ships (single-EPP and 3-EPP deployments), and drops the coordinator's private copy of the gateway manifests.
Use the shared inference-gateway component. The suite loads EPP, InferencePool, RBAC, ServiceAccounts, and Services from
deploy/components/inference-gateway/instead of the coordinator-localepd-pool/manifests. As in the router e2e, the coordinator fronts the EPP with a hand-rolled Envoy, so that component'sgateways.yaml/httproutes.yamlstay unused.Consolidate the coordinator manifests. Flattens
deploy/coordinator/components/coordinator/up todeploy/coordinator/and deletes the now-redundant manifests it replaced.Add an optional 3-EPP topology (
E2E_EPP_TOPOLOGY=3epp). Runs one role-scoped EPP and InferencePool per phase (encode / prefill / decode), with an Envoy route that dispatches eachEPP-Profileto the matching role. Each EPP sees only its own role's pods, so it usessingle-profile-handlerand no role filter, with per-role scorers mirroring the guide. The default single-EPP topology is unchanged.Extract common resources.
shared-envoy-resources.yamlholds the Envoy Deployment and Service; per-topology manifests carry only the routing ConfigMap. Topology branching lives ineppsToCreate()/poolNames().Deploy the renderer like the router e2e.
vllm-renderruns as a standalone Deployment/Service, reached through the substituted${VLLM_RENDER_URL}.CI.
e2e-testsbecomes atopology: [single, 3epp]matrix, aggregated by ane2e-tests-statusjob (required checks can't reference individual matrix legs).Follow-up: a later PR will add parallelization to coordinator e2e tests, as in the router.
Which issue(s) this PR fixes:
Fixes #
Release note (write
NONEif no user-facing change):