Remove deprecated inline telemetry field from CRDs

[ChrisJBurns]

Summary

The inline telemetry field on MCPServerSpec and MCPRemoteProxySpec was deprecated in favor of telemetryConfigRef, which references a shared MCPTelemetryConfig resource. Similarly, the config.telemetry/telemetryConfigRef mutual-exclusivity CEL rule on VirtualMCPServer existed only to bridge the migration period. This removes all backward-compatibility shim code, CEL rules, deprecated utility functions, and the TelemetryConfig/OpenTelemetryConfig CRD types ahead of the v1beta1 API promotion.

Additionally, the operator-side converter no longer falls back to config.telemetry for VirtualMCPServer (D4). Operator-managed vMCPs must use telemetryConfigRef.

Changes

File	Change
api/v1alpha1/mcpserver_types.go	Remove Telemetry field, CEL rule, TelemetryConfig/OpenTelemetryConfig types; keep shared sub-types
api/v1alpha1/mcpremoteproxy_types.go	Remove Telemetry field, CEL rule
api/v1alpha1/virtualmcpserver_types.go	Remove CEL rule for config.telemetry/telemetryConfigRef
controllers/mcpserver_runconfig.go	Remove else branch calling deprecated AddTelemetryConfigOptions
controllers/mcpserver_controller.go	Remove 2 deprecated env var generation paths
controllers/mcpremoteproxy_runconfig.go	Simplify addTelemetryOptions to only handle TelemetryConfigRef
controllers/mcpremoteproxy_deployment.go	Remove deprecated inline telemetry env var path
pkg/runconfig/telemetry.go	Remove AddTelemetryConfigOptions function
pkg/spectoconfig/telemetry.go	Remove ConvertTelemetryConfig function
pkg/controllerutil/tokenexchange.go	Remove GenerateOpenTelemetryEnvVars function
pkg/vmcpconfig/converter.go	Remove deprecated inline fallback in normalizeTelemetry; operator now returns nil when TelemetryConfigRef is not set
pkg/export/k8s.go	Remove inline telemetry export (requires separate MCPTelemetryConfig now)
cmd/thv/app/export.go	Add stderr warning when telemetry config is detected during export
*_test.go, test-integration/	Remove/update all tests for deprecated code paths
deploy/charts/operator-crds/	Regenerated CRD YAML (inline telemetry schema removed)

Type of change

• Refactoring (no behavior change)

Test plan

• Unit tests (task operator-test)
• Linting (task lint-fix)
• Build (go build ./...)

Does this introduce a user-facing change?

Yes — this is a breaking change for existing CRs. See migration guide below.

Migration guide

MCPServer and MCPRemoteProxy

Before (inline telemetry — no longer supported):

spec:
  telemetry:
    openTelemetry:
      enabled: true
      endpoint: "otel-collector:4317"
      tracing:
        enabled: true
    prometheus:
      enabled: true

After (shared MCPTelemetryConfig resource):

# 1. Create a shared MCPTelemetryConfig resource
apiVersion: toolhive.stacklok.dev/v1alpha1
kind: MCPTelemetryConfig
metadata:
  name: shared-otel
spec:
  openTelemetry:
    enabled: true
    endpoint: "otel-collector:4317"
    tracing:
      enabled: true
  prometheus:
    enabled: true
---
# 2. Reference it from your MCPServer/MCPRemoteProxy
spec:
  telemetryConfigRef:
    name: shared-otel
    serviceName: my-server  # per-server override (optional)

VirtualMCPServer

config.telemetry is no longer read by the operator. Use spec.telemetryConfigRef instead (same MCPTelemetryConfig pattern as above).

thv export

thv export --format k8s now prints a stderr warning when telemetry config is detected but cannot be exported inline. Users must create an MCPTelemetryConfig resource manually.

Large PR Justification

• we're removing entire portions of code

Special notes for reviewers

• Shared types kept: PrometheusConfig, OpenTelemetryTracingConfig, and OpenTelemetryMetricsConfig are still used by MCPTelemetryConfig and were intentionally retained.
• Config.Telemetry in shared struct: The Telemetry field in pkg/vmcp/config/config.go is still valid for standalone CLI deployments — only the operator-side fallback was removed.
• This is PR 1 of 5 in the deprecation removal series before v1beta1 promotion.

Closes #4828

Generated with Claude Code

[github-actions]

Large PR Detected

This PR exceeds 1000 lines of changes and requires justification before it can be reviewed.

How to unblock this PR:

Add a section to your PR description with the following format:

## Large PR Justification

[Explain why this PR must be large, such as:]
- Generated code that cannot be split
- Large refactoring that must be atomic
- Multiple related changes that would break if separated
- Migration or data transformation

Alternative:

Consider splitting this PR into smaller, focused changes (< 1000 lines each) for easier review and reduced risk.

See our Contributing Guidelines for more details.

---

This review will be automatically dismissed once you add the justification section.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 68.97%. Comparing base (42f7395) to head (e940bbd).

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #4819      +/-   ##
==========================================
- Coverage   69.10%   68.97%   -0.13%     
==========================================
  Files         530      530              
  Lines       55379    55281      -98     
==========================================
- Hits        38267    38128     -139     
- Misses      14170    14212      +42     
+ Partials     2942     2941       -1     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[ChrisJBurns]

Multi-Agent Consensus Review

Agents consulted: kubernetes-expert, code-reviewer, toolhive-expert, site-reliability-engineer

Consensus Summary

#	Finding	Consensus	Severity	Action
1	Stale comments referencing removed types/functions	10/10	LOW	Fix
2	Architecture docs outdated	8/10	MEDIUM	Fix
3	thv export silently drops telemetry config	8/10	MEDIUM	Discuss
4	Example YAML files have stale deprecation comments	8/10	LOW	Fix
5	MCPRemoteProxy pre-existing env var gap	7/10	INFO	Note

Overall

Clean, thorough deprecation removal. The structural changes are correct: field and type removals, deepcopy updates, CEL rule removals, controller fallback branch removals, CRD YAML regeneration, and test cleanup are all internally consistent across MCPServer, MCPRemoteProxy, and VirtualMCPServer. No compilation risks, nil pointer issues, or dead code paths found. Shared types (PrometheusConfig, OpenTelemetryTracingConfig, OpenTelemetryMetricsConfig) are correctly retained for MCPTelemetryConfig.

All five findings are non-blocking. The two MEDIUM findings (architecture doc drift and silent telemetry drop in thv export) are polish items that could be addressed in this PR or in a follow-up. The stale comment fixes are straightforward search-and-replace. The MCPRemoteProxy env var gap is pre-existing and out of scope for this PR.

File-level findings (not on diff lines)

[LOW] F1: Stale comments referencing removed types/functions (Consensus: 10/10)

Multiple comments still reference removed types, deleted functions, or stale phrases:

• mcpserver_controller.go:1024,1634 — "prefer TelemetryConfigRef over deprecated inline"
• mcpremoteproxy_runconfig.go:112 — "prefer TelemetryConfigRef over deprecated inline Telemetry"
• mcptelemetryconfig_types.go:36 — references deleted OpenTelemetryConfig type
• converter_test.go:1484,1618 — references deleted ConvertTelemetryConfig
• converter.go:126-127 — references CEL mutual exclusion that was removed

Raised by: k8s-reviewer, code-reviewer, toolhive-reviewer, sre-reviewer

[INFO] F5: Pre-existing gap — MCPRemoteProxy missing GenerateOpenTelemetryEnvVarsFromRef (Consensus: 7/10)

Unlike MCPServer and VirtualMCPServer, MCPRemoteProxy never calls GenerateOpenTelemetryEnvVarsFromRef to inject OTEL_RESOURCE_ATTRIBUTES into pods. The removed GenerateOpenTelemetryEnvVars was the only env var injection path. This is pre-existing — not a regression from this PR — but worth noting for a follow-up.

Raised by: sre-reviewer

Documentation

• docs/arch/09-operator-architecture.md lines 161, 163, 662-664 still reference the now-removed inline telemetry field and CEL mutual-exclusivity rules. Per CLAUDE.md: "When making changes that affect architecture, update relevant docs in docs/arch/."
• examples/operator/mcp-servers/mcpserver_fetch_otel.yaml:25 and examples/operator/virtual-mcps/vmcp_with_telemetry_ref.yaml:79 describe the field as "deprecated" rather than "removed."

---

Generated with Claude Code

[github-actions]

✅ Large PR justification has been provided. The size review has been dismissed and this PR can now proceed with normal review.

Large PR justification has been provided. Thank you!

[github-actions]

Large PR Detected

This PR exceeds 1000 lines of changes and requires justification before it can be reviewed.

How to unblock this PR:

Add a section to your PR description with the following format:

## Large PR Justification

[Explain why this PR must be large, such as:]
- Generated code that cannot be split
- Large refactoring that must be atomic
- Multiple related changes that would break if separated
- Migration or data transformation

Alternative:

Consider splitting this PR into smaller, focused changes (< 1000 lines each) for easier review and reduced risk.

See our Contributing Guidelines for more details.

---

This review will be automatically dismissed once you add the justification section.

Large PR justification has been provided. Thank you!

[github-actions]

✅ Large PR justification has been provided. The size review has been dismissed and this PR can now proceed with normal review.