docs(mcp): note Istio strict-mTLS incompatibility for Service backendRefs - #2430
Open
Aias00 wants to merge 1 commit into
Open
docs(mcp): note Istio strict-mTLS incompatibility for Service backendRefs#2430Aias00 wants to merge 1 commit into
Aias00 wants to merge 1 commit into
Conversation
…Refs An MCPRoute backendRef can reference an Envoy Gateway Backend (FQDN) or a Kubernetes Service. When it references an in-cluster Service, Envoy resolves it via EndpointSlices — i.e. connects directly to pod IPs, not the Service cluster IP — which bypasses strict-mTLS meshes (e.g. Istio PeerAuthentication STRICT + DestinationRule) that enforce TLS at the service level. The symptom is a failed MCP session with no AI Gateway error. This is expected Envoy/Gateway-API Service resolution, not an AI Gateway bug, but it was not documented. Add a subsection to the MCPRoute docs explaining the behavior and the workarounds (prefer a Backend with the Service FQDN, relax mTLS for the backend port, or exclude the port/workload from the mesh). Addresses envoyproxy#2003 Co-Authored-By: Claude <noreply@anthropic.com> Signed-off-by: liuhy <liuhongyu@apache.org>
dosubot
Bot
added
the
size:M
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Description
An
MCPRoutebackendRefcan reference either an Envoy GatewayBackend(FQDN endpoint) or a KubernetesService. When it references an in-clusterService, Envoy resolves it viaEndpointSlices— i.e. it connects directly to individual pod IPs, not to the Service cluster IP. A strict-mTLS mesh (e.g. IstioPeerAuthenticationSTRICT +DestinationRule) enforces TLS at the service level, so the pod-IP connections bypass it and the MCP session fails to establish, with no error surfaced by the AI Gateway.This is expected Envoy/Gateway-API
Serviceresolution, not an AI Gateway bug, but it was not reflected anywhere in the documentation (the original ask in #2003). This adds a subsection to the MCPRoute docs explaining the behavior and the workarounds:Backendwith the Service FQDN (<service>.<namespace>.svc.cluster.local) over aServicebackendRef — resolves via STRICT_DNS to the Service cluster IP, so the mesh applies strict mTLS at the service level. (Example YAML included.)portLevelMtls.mode: PERMISSIVE).This is a docs-only change; the underlying Service→EndpointSlice resolution is Envoy/Gateway-API behavior and an architectural change to it would be a separate, maintainer-level decision.
Related Issues/PRs (if applicable)
Addresses #2003
Special notes for reviewers (if applicable)
Backenddocs link uses the same anchor convention as existing links in the repo (e.g.#ratelimitselectcondition,#oidc).ServicebackendRefs to a STRICT_DNS / Service-FQDN cluster in the extension server) is desired, that is a larger, separately-verifiable change and out of scope for this docs PR.