You are viewing a plain text version of this content. The canonical link for it is here.
Posted to notifications@skywalking.apache.org by wa...@apache.org on 2026/09/22 05:42:12 UTC

(skywalking-showcase) 01/01: Send the OpenTelemetry copy's traces straight to the OAP, with no collector

This is an automated email from the ASF dual-hosted git repository.

wankai123 pushed a commit to branch feature/otlp-trace-direct
in repository https://gitbox.apache.org/repos/asf/skywalking-showcase.git

commit 618c1bc2e2bc13b578a088a1fa3cb448b4dfdecd
Author: wankai123 <wa...@foxmail.com>
AuthorDate: Tue Sep 22 11:16:31 2026 +0800

    Send the OpenTelemetry copy's traces straight to the OAP, with no collector
    
    The OAP serves OTLP on its own ports, gRPC on 11800 and OTLP/HTTP on 12800, so the second
    copy of the music application no longer needs a collector to relay its spans. Each sender
    uses the transport its instrumentation has: the Java and Node.js agents speak gRPC, while
    the Python auto-instrumentation image ships no gRPC exporter and the APISIX plugin has no
    gRPC client at all, so those two use OTLP/HTTP. On Kubernetes the collector keeps only the
    Prometheus pipelines the monitoring features need, and feature-otlp-trace no longer turns
    it on; on Docker Compose the copy's own collector is gone and the OAP joins the copy's
    network instead, which keeps the two copies isolated because their services answer to the
    same hostnames.
    
    The Zipkin store becomes the mesh's alone, fed by Istio's sidecars, so /zipkin shows the
    mesh and /otlp this copy rather than both showing the same spans under two sets of service
    names. The Docker layout runs no mesh, so its Zipkin store stays empty; the receiver and
    the query stay on there only so the provisioned ZipkinTraceQL datasource and its dashboard
    answer instead of 404.
    
    The OAP image moves to apache/skywalking#14100, which this design needs: a collector used
    to mask the fact that APISIX sends successful spans without a status, and Grafana's trace
    view crashes on such a span until the OAP fills the field in.
---
 Makefile.in                                        |  2 +-
 .../docker/config/apisix-otlp-trace/config.yaml    |  2 +-
 .../config/otel-collector-config-otlp-trace.yaml   | 41 --------------
 deploy/platform/docker/docker-compose.cluster.yaml |  8 +--
 .../platform/docker/docker-compose.otlp-trace.yaml | 63 +++++++++++-----------
 .../docker/docker-compose.single-node.yaml         |  8 +--
 deploy/platform/kubernetes/Makefile                |  1 -
 .../templates/feature-otlp-trace/resources.yaml    | 28 ++++++----
 .../templates/otel-collector-config.yaml           | 24 ---------
 deploy/platform/kubernetes/values.yaml             |  6 +--
 docs/readme.md                                     |  2 +-
 11 files changed, 64 insertions(+), 121 deletions(-)

diff --git a/Makefile.in b/Makefile.in
index 8047382..6f85374 100644
--- a/Makefile.in
+++ b/Makefile.in
@@ -38,7 +38,7 @@ ES_IMAGE_TAG ?= 7.10.2
 
 SW_OAP_IMAGE ?= ghcr.io/apache/skywalking/oap
 # banyandb-tail-sampling needs the bydb.yml `pipeline` block, which lands in c66a476a (#13965).
-SW_OAP_IMAGE_TAG ?= ecb087cbfd23a89581367ca70adc74f511a9e70c
+SW_OAP_IMAGE_TAG ?= 03ecf41fd7ac34635370c91e2bf01e0424fa85eb
 
 SW_UI_IMAGE ?= ghcr.io/apache/skywalking-horizon-ui
 SW_UI_IMAGE_TAG ?= dc31135cadee387263931f359819e7a8c7971bdc
diff --git a/deploy/platform/docker/config/apisix-otlp-trace/config.yaml b/deploy/platform/docker/config/apisix-otlp-trace/config.yaml
index e91ad79..435f7a6 100644
--- a/deploy/platform/docker/config/apisix-otlp-trace/config.yaml
+++ b/deploy/platform/docker/config/apisix-otlp-trace/config.yaml
@@ -26,7 +26,7 @@ plugin_attr:
       service.name: frontend
       service.namespace: otel-services
     collector:
-      address: otel-trace-collector:4318
+      address: oap:12800
       request_timeout: 3
 plugins:
   - opentelemetry
diff --git a/deploy/platform/docker/config/otel-collector-config-otlp-trace.yaml b/deploy/platform/docker/config/otel-collector-config-otlp-trace.yaml
deleted file mode 100644
index 9547f81..0000000
--- a/deploy/platform/docker/config/otel-collector-config-otlp-trace.yaml
+++ /dev/null
@@ -1,41 +0,0 @@
-# Licensed to the Apache Software Foundation (ASF) under one
-# or more contributor license agreements.  See the NOTICE file
-# distributed with this work for additional information
-# regarding copyright ownership.  The ASF licenses this file
-# to you under the Apache License, Version 2.0 (the
-# "License"); you may not use this file except in compliance
-# with the License.  You may obtain a copy of the License at
-#
-#   http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing,
-# software distributed under the License is distributed on an
-# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
-# KIND, either express or implied.  See the License for the
-# specific language governing permissions and limitations
-# under the License.
-#
-# @feature: otlp-trace; receives the spans of the OpenTelemetry-instrumented copy of the music application, over
-# OTLP/HTTP, the SDKs' default protocol, and hands
-# every one to the OAP twice: as OTLP over gRPC, stored natively, and converted to Zipkin for the Zipkin store.
-receivers:
-  otlp:
-    protocols:
-      grpc:
-        endpoint: 0.0.0.0:4317
-      http:
-        endpoint: 0.0.0.0:4318
-
-exporters:
-  otlp:
-    endpoint: oap:11800
-    tls:
-      insecure: true
-  zipkin:
-    endpoint: http://oap:9411/api/v2/spans
-
-service:
-  pipelines:
-    traces:
-      receivers: [ otlp ]
-      exporters: [ otlp, zipkin ]
diff --git a/deploy/platform/docker/docker-compose.cluster.yaml b/deploy/platform/docker/docker-compose.cluster.yaml
index 6946f6b..4f4b2d3 100644
--- a/deploy/platform/docker/docker-compose.cluster.yaml
+++ b/deploy/platform/docker/docker-compose.cluster.yaml
@@ -61,9 +61,11 @@ services:
       SW_TRACEQL: default # @feature: grafana; serve the Tempo API (TraceQL) that the Grafana trace dashboards query
       SW_TRACEQL_ENABLE_DATASOURCE_SKYWALKING: "true" # @feature: grafana; the /skywalking datasource, traces from the SkyWalking agents
       SW_TRACEQL_ENABLE_DATASOURCE_OTLP: "true" # @feature: otlp-trace,grafana; the /otlp datasource, OTLP traces stored natively (receiver-otel.otlpTraceStorage: otlp, the default)
-      SW_RECEIVER_ZIPKIN: default # @feature: otlp-trace; the collector hands every span of the OpenTelemetry-instrumented copy to the Zipkin store too
-      SW_QUERY_ZIPKIN: default # @feature: otlp-trace,grafana; the Zipkin query API and the /zipkin datasource
-      SW_TRACEQL_ENABLE_DATASOURCE_ZIPKIN: "true" # @feature: otlp-trace,grafana; the /zipkin datasource
+      # The Zipkin store is fed by Istio's sidecars, which only the Kubernetes layout runs, so here it stays empty;
+      # the receiver and the query are on so the ZipkinTraceQL datasource and its dashboard answer rather than 404.
+      SW_RECEIVER_ZIPKIN: default # @feature: grafana; the Zipkin receiver on 9411
+      SW_QUERY_ZIPKIN: default # @feature: grafana; the Zipkin query API and the /zipkin datasource
+      SW_TRACEQL_ENABLE_DATASOURCE_ZIPKIN: "true" # @feature: grafana; the /zipkin datasource
       JAVA_OPTS: "-Dmode=no-init -Xms2048m -Xmx2048m" # @feature: cluster; start the cluster nodes in no-init mode
       SW_METER_ANALYZER_ACTIVE_FILES: datasource,threadpool,satellite,spring-micrometer,go-runtime,java-agent,go-agent
     healthcheck:
diff --git a/deploy/platform/docker/docker-compose.otlp-trace.yaml b/deploy/platform/docker/docker-compose.otlp-trace.yaml
index 15521cf..7d85fcb 100644
--- a/deploy/platform/docker/docker-compose.otlp-trace.yaml
+++ b/deploy/platform/docker/docker-compose.otlp-trace.yaml
@@ -16,36 +16,35 @@
 # under the License.
 #
 # @feature: otlp-trace; a second copy of the music application from the agentless images, instrumented at start by the
-# OpenTelemetry auto-instrumentation (copied out of the operator's images into shared volumes) and exporting OTLP to
-# its own collector, which hands every span to the OAP as OTLP and as Zipkin. The copy lives on its own network, where
-# the aliases give the services the fixed hostnames they address each other by: gateway, songs, rcmd, rating, app,
-# activemq. The Go rating service has no runtime instrumentation and appears only as recommendation's callee.
+# OpenTelemetry auto-instrumentation (copied out of the operator's images into shared volumes) and exporting every span
+# straight to the OAP, which stores it natively and serves it at the TraceQL /otlp datasource. The copy lives on its own
+# network, where the aliases give the services the fixed hostnames they address each other by: frontend, gateway, songs,
+# rcmd, rating, app, activemq. The Go rating service has no runtime instrumentation and appears only as
+# recommendation's callee.
 
 version: '2.1'
 
+# Every span goes straight to the OAP; no collector sits in the trace path. The OAP serves OTLP on both of its
+# own ports, gRPC on 11800 and OTLP/HTTP on 12800, so each sender uses the transport its instrumentation has.
 x-otel-export: &otel-export
-  OTEL_EXPORTER_OTLP_PROTOCOL: http/protobuf
-  OTEL_EXPORTER_OTLP_ENDPOINT: http://otel-trace-collector:4318
   OTEL_TRACES_EXPORTER: otlp
   OTEL_METRICS_EXPORTER: none
   OTEL_LOGS_EXPORTER: none
   OTEL_RESOURCE_ATTRIBUTES: service.namespace=otel-services
+x-otel-grpc: &otel-grpc
+  OTEL_EXPORTER_OTLP_PROTOCOL: grpc
+  OTEL_EXPORTER_OTLP_ENDPOINT: http://oap:11800
+# The Python auto-instrumentation image carries no gRPC exporter, so it goes over OTLP/HTTP.
+x-otel-http: &otel-http
+  OTEL_EXPORTER_OTLP_PROTOCOL: http/protobuf
+  OTEL_EXPORTER_OTLP_ENDPOINT: http://oap:12800
 
 services:
-  otel-trace-collector:
-    image: ${OTEL_COLLECTOR_IMAGE}:${OTEL_COLLECTOR_IMAGE_TAG}
+  # The OAP joins this copy's network so the services can reach it by name. Only the OAP crosses over: the copy
+  # keeps its own network because its services answer to the same hostnames as the agent copy on `sw`.
+  oap:
     networks: [ sw, otel ]
-    command: [ "--config=/etc/otel-collector-config.yaml" ]
-    volumes:
-      - ./config/otel-collector-config-otlp-trace.yaml:/etc/otel-collector-config.yaml
-    expose:
-      - 4317
-      - 4318
-    depends_on:
-      oap:
-        condition: service_healthy
 
-  # One-shot copies of the auto-instrumentation into the volumes the services mount.
   otel-instrumentation-java:
     image: ${OTEL_INSTRUMENTATION_JAVA_IMAGE}
     command: [ "cp", "/javaagent.jar", "/otel/javaagent.jar" ]
@@ -69,7 +68,7 @@ services:
       otel:
         aliases: [ gateway ]
     environment:
-      <<: *otel-export
+      <<: [ *otel-export, *otel-grpc ]
       OTEL_SERVICE_NAME: gateway
       JAVA_TOOL_OPTIONS: -javaagent:/otel/javaagent.jar
     volumes:
@@ -81,8 +80,8 @@ services:
     depends_on:
       otel-instrumentation-java:
         condition: service_completed_successfully
-      otel-trace-collector:
-        condition: service_started
+      oap:
+        condition: service_healthy
 
   otel-songs:
     restart: unless-stopped
@@ -91,7 +90,7 @@ services:
       otel:
         aliases: [ songs ]
     environment:
-      <<: *otel-export
+      <<: [ *otel-export, *otel-grpc ]
       OTEL_SERVICE_NAME: songs
       JAVA_TOOL_OPTIONS: -javaagent:/otel/javaagent.jar
       ACTIVE_MQ_URL: tcp://activemq:61616
@@ -105,8 +104,8 @@ services:
     depends_on:
       otel-instrumentation-java:
         condition: service_completed_successfully
-      otel-trace-collector:
-        condition: service_started
+      oap:
+        condition: service_healthy
       otel-activemq:
         condition: service_started
 
@@ -117,7 +116,7 @@ services:
       otel:
         aliases: [ rcmd ]
     environment:
-      <<: *otel-export
+      <<: [ *otel-export, *otel-http ]
       OTEL_SERVICE_NAME: recommendation
       PYTHONPATH: /otel/opentelemetry/instrumentation/auto_instrumentation:/otel
     volumes:
@@ -129,8 +128,8 @@ services:
     depends_on:
       otel-instrumentation-python:
         condition: service_completed_successfully
-      otel-trace-collector:
-        condition: service_started
+      oap:
+        condition: service_healthy
 
   otel-app:
     restart: unless-stopped
@@ -139,7 +138,7 @@ services:
       otel:
         aliases: [ app ]
     environment:
-      <<: *otel-export
+      <<: [ *otel-export, *otel-grpc ]
       OTEL_SERVICE_NAME: app
       NODE_OPTIONS: --require /otel/autoinstrumentation.js
       OTEL_NODE_RESOURCE_DETECTORS: env,host,os,process,serviceinstance
@@ -157,8 +156,8 @@ services:
     depends_on:
       otel-instrumentation-nodejs:
         condition: service_completed_successfully
-      otel-trace-collector:
-        condition: service_started
+      oap:
+        condition: service_healthy
 
   otel-rating:
     restart: unless-stopped
@@ -202,8 +201,8 @@ services:
     depends_on:
       otel-app:
         condition: service_healthy
-      otel-trace-collector:
-        condition: service_started
+      oap:
+        condition: service_healthy
   # The same Selenium load generator as the agent copy, driving the frontend page.
   otel-loadgen:
     restart: unless-stopped
diff --git a/deploy/platform/docker/docker-compose.single-node.yaml b/deploy/platform/docker/docker-compose.single-node.yaml
index 230c4fd..b43b7af 100644
--- a/deploy/platform/docker/docker-compose.single-node.yaml
+++ b/deploy/platform/docker/docker-compose.single-node.yaml
@@ -54,9 +54,11 @@ services:
       SW_TRACEQL: default # @feature: grafana; serve the Tempo API (TraceQL) that the Grafana trace dashboards query
       SW_TRACEQL_ENABLE_DATASOURCE_SKYWALKING: "true" # @feature: grafana; the /skywalking datasource, traces from the SkyWalking agents
       SW_TRACEQL_ENABLE_DATASOURCE_OTLP: "true" # @feature: otlp-trace,grafana; the /otlp datasource, OTLP traces stored natively (receiver-otel.otlpTraceStorage: otlp, the default)
-      SW_RECEIVER_ZIPKIN: default # @feature: otlp-trace; the collector hands every span of the OpenTelemetry-instrumented copy to the Zipkin store too
-      SW_QUERY_ZIPKIN: default # @feature: otlp-trace,grafana; the Zipkin query API and the /zipkin datasource
-      SW_TRACEQL_ENABLE_DATASOURCE_ZIPKIN: "true" # @feature: otlp-trace,grafana; the /zipkin datasource
+      # The Zipkin store is fed by Istio's sidecars, which only the Kubernetes layout runs, so here it stays empty;
+      # the receiver and the query are on so the ZipkinTraceQL datasource and its dashboard answer rather than 404.
+      SW_RECEIVER_ZIPKIN: default # @feature: grafana; the Zipkin receiver on 9411
+      SW_QUERY_ZIPKIN: default # @feature: grafana; the Zipkin query API and the /zipkin datasource
+      SW_TRACEQL_ENABLE_DATASOURCE_ZIPKIN: "true" # @feature: grafana; the /zipkin datasource
       JAVA_OPTS: "-Xms2048m -Xmx2048m"
       SW_METER_ANALYZER_ACTIVE_FILES: datasource,threadpool,satellite,spring-micrometer,go-runtime,java-agent,go-agent
     healthcheck:
diff --git a/deploy/platform/kubernetes/Makefile b/deploy/platform/kubernetes/Makefile
index b3a87d1..2cf4d82 100644
--- a/deploy/platform/kubernetes/Makefile
+++ b/deploy/platform/kubernetes/Makefile
@@ -270,7 +270,6 @@ feature-grafana:
 .PHONY: feature-otlp-trace
 feature-otlp-trace:
 	$(eval HELM_OPTIONS := $(HELM_OPTIONS) --set features.otlpTrace.enabled=true)
-	$(eval HELM_OPTIONS := $(HELM_OPTIONS) --set opentelemetry.enabled=true)
 	$(eval HELM_OPTIONS := $(HELM_OPTIONS) --set features.otlpTrace.instrumentation.java=$(OTEL_INSTRUMENTATION_JAVA_IMAGE))
 	$(eval HELM_OPTIONS := $(HELM_OPTIONS) --set features.otlpTrace.instrumentation.nodejs=$(OTEL_INSTRUMENTATION_NODEJS_IMAGE))
 	$(eval HELM_OPTIONS := $(HELM_OPTIONS) --set features.otlpTrace.instrumentation.python=$(OTEL_INSTRUMENTATION_PYTHON_IMAGE))
diff --git a/deploy/platform/kubernetes/templates/feature-otlp-trace/resources.yaml b/deploy/platform/kubernetes/templates/feature-otlp-trace/resources.yaml
index 69345ee..6c78c3e 100644
--- a/deploy/platform/kubernetes/templates/feature-otlp-trace/resources.yaml
+++ b/deploy/platform/kubernetes/templates/feature-otlp-trace/resources.yaml
@@ -19,14 +19,19 @@
 {{- $ns := .Values.features.otlpTrace.namespace }}
 {{- $hub := .Values.sampleServices.hub }}
 {{- $tag := .Values.sampleServices.tag }}
-{{- $collector := printf "%s-opentelemetry-collector.%s.svc.cluster.local" .Release.Name .Release.Namespace }}
+{{- /* The spans go straight to the OAP: gRPC on 11800 where the instrumentation speaks it, OTLP/HTTP on 12800
+       for the Python agent and the APISIX plugin, which are HTTP-only. */}}
+{{- $oapGrpc := include "skywalking.oap.address.grpc" . }}
+{{- $oapHttp := include "skywalking.oap.address.http" . }}
+{{- $oapHost := include "skywalking.oap.address.host" . }}
 {{- /*
 A second copy of the music application: the agentless images, the OpenTelemetry auto-instrumentation copied in by an
 init container and switched on through the runtime's own hook (JAVA_TOOL_OPTIONS, NODE_OPTIONS, PYTHONPATH), the APISIX
-frontend traced by its built-in opentelemetry plugin, and the spans exported over OTLP/HTTP, the SDKs' default, to the
-OpenTelemetry Collector, which hands them to the OAP as OTLP and as Zipkin. The same Selenium load generator drives the
-frontend page, so the traces start at frontend. It has its own namespace because the services address each other by
-fixed hostnames, frontend, gateway, songs, rcmd, rating, app and activemq.
+frontend traced by its built-in opentelemetry plugin, and every span exported straight to the OAP, which stores it
+natively for the TraceQL /otlp datasource. No collector sits in the trace path. The Zipkin store stays the mesh's:
+Istio's sidecars send it their own spans, so /zipkin shows the mesh and /otlp this copy.
+The same Selenium load generator drives the frontend page, so the traces start at frontend. It has its own namespace
+because the services address each other by fixed hostnames, frontend, gateway, songs, rcmd, rating, app and activemq.
 The page carries no browser agent, and the Go rating service has no runtime instrumentation and appears only as the
 callee of recommendation's spans.
 */}}
@@ -141,9 +146,9 @@ spec:
             - name: OTEL_RESOURCE_ATTRIBUTES
               value: service.namespace={{ $ns }}
             - name: OTEL_EXPORTER_OTLP_PROTOCOL
-              value: http/protobuf
+              value: grpc
             - name: OTEL_EXPORTER_OTLP_ENDPOINT
-              value: http://{{ $collector }}:4318
+              value: http://{{ $oapGrpc }}
             - name: OTEL_TRACES_EXPORTER
               value: otlp
             - name: OTEL_METRICS_EXPORTER
@@ -216,10 +221,11 @@ spec:
               value: recommendation
             - name: OTEL_RESOURCE_ATTRIBUTES
               value: service.namespace={{ $ns }}
+            # The Python auto-instrumentation image carries no gRPC exporter, so this one goes over OTLP/HTTP.
             - name: OTEL_EXPORTER_OTLP_PROTOCOL
               value: http/protobuf
             - name: OTEL_EXPORTER_OTLP_ENDPOINT
-              value: http://{{ $collector }}:4318
+              value: {{ $oapHttp }}
             - name: OTEL_TRACES_EXPORTER
               value: otlp
             - name: OTEL_METRICS_EXPORTER
@@ -287,9 +293,9 @@ spec:
             - name: OTEL_RESOURCE_ATTRIBUTES
               value: service.namespace={{ $ns }}
             - name: OTEL_EXPORTER_OTLP_PROTOCOL
-              value: http/protobuf
+              value: grpc
             - name: OTEL_EXPORTER_OTLP_ENDPOINT
-              value: http://{{ $collector }}:4318
+              value: http://{{ $oapGrpc }}
             - name: OTEL_TRACES_EXPORTER
               value: otlp
             - name: OTEL_METRICS_EXPORTER
@@ -372,7 +378,7 @@ data:
           service.name: frontend
           service.namespace: {{ $ns }}
         collector:
-          address: {{ $collector }}:4318
+          address: {{ $oapHost }}:12800
           request_timeout: 3
     plugins:
       - opentelemetry
diff --git a/deploy/platform/kubernetes/templates/otel-collector-config.yaml b/deploy/platform/kubernetes/templates/otel-collector-config.yaml
index 2d17748..46b40f3 100644
--- a/deploy/platform/kubernetes/templates/otel-collector-config.yaml
+++ b/deploy/platform/kubernetes/templates/otel-collector-config.yaml
@@ -42,16 +42,6 @@ metadata:
 data:
   config.yaml: |
     receivers:
-      {{- if .Values.features.otlpTrace.enabled }}
-      # @feature: otlp-trace; the OpenTelemetry-instrumented copy of the music application exports its spans here over
-      # OTLP/HTTP, the SDKs' default; gRPC stays open for anything else pointed at the collector
-      otlp:
-        protocols:
-          grpc:
-            endpoint: 0.0.0.0:4317
-          http:
-            endpoint: 0.0.0.0:4318
-      {{- end }}
       {{- $root := . -}}
       {{- range $idx, $config := $configs -}}
       {{- if not (include $config $root | trim | empty) }}
@@ -72,15 +62,6 @@ data:
           insecure: true
       {{- end }}
       {{- end }}
-      {{- if .Values.features.otlpTrace.enabled }}
-      # @feature: otlp-trace; the same spans go to the OAP twice: natively as OTLP, and as Zipkin for the Zipkin store
-      otlp/traces:
-        endpoint: "{{ include "skywalking.oap.address.grpc" . }}"
-        tls:
-          insecure: true
-      zipkin/traces:
-        endpoint: "http://{{ include "skywalking.oap.address.host" . }}:9411/api/v2/spans"
-      {{- end }}
       logging:
         verbosity: detailed
 
@@ -92,11 +73,6 @@ data:
 
     service:
       pipelines:
-      {{- if .Values.features.otlpTrace.enabled }}
-        traces:
-          receivers: [ otlp ]
-          exporters: [ otlp/traces, zipkin/traces ]
-      {{- end }}
       {{- $root := . }}
       {{- range $idx, $config := $configs }}
       {{- if not (include $config $root | trim | empty) }}
diff --git a/deploy/platform/kubernetes/values.yaml b/deploy/platform/kubernetes/values.yaml
index a454432..83fb626 100644
--- a/deploy/platform/kubernetes/values.yaml
+++ b/deploy/platform/kubernetes/values.yaml
@@ -929,9 +929,9 @@ features:
     enabled: false
     image: ~
   # @feature: otlp-trace; a second copy of the music application, from the agentless images, instrumented at start by the
-  # OpenTelemetry auto-instrumentation (the APISIX frontend by its own opentelemetry plugin) and exporting OTLP to the
-  # OpenTelemetry Collector, which hands every span to the OAP as OTLP, stored natively for Grafana's /otlp, and as Zipkin
-  # for the Zipkin store and /zipkin
+  # OpenTelemetry auto-instrumentation (the APISIX frontend by its own opentelemetry plugin) and exporting every span
+  # straight to the OAP, gRPC on 11800 or OTLP/HTTP on 12800, stored natively for Grafana's /otlp. No collector sits in
+  # the trace path; the Zipkin store stays the mesh's.
   otlpTrace:
     enabled: false
     # Its own namespace: the services address each other by the same fixed hostnames as the agent copy.
diff --git a/docs/readme.md b/docs/readme.md
index e67dc7b..ebd1b5b 100644
--- a/docs/readme.md
+++ b/docs/readme.md
@@ -115,7 +115,7 @@ Currently, the features supported are:
 | `apisix-monitor`        | Deploy OpenTelemetry and export APISIX metrics to SkyWalking for analysis and display on UI                                                                                            |                                                                                                                                       |
 | `mesh-with-agent`       | Deploy services with java agent in the service mesh environment.                                                                                                                       | Only support deployment in the Kubernetes environment, docker is not supported.                                                       |
 | `grafana`               | Deploy a Grafana to show SkyWalking metrics and logs on the Grafana UI.                                                                                                                | Feel free to modify the Grafana config when deploy your own environment.                                                              |
-| `otlp-trace`            | Deploy a second copy of the music application from its agentless images: the Java, Node.js and Python services instrumented at start by the OpenTelemetry auto-instrumentation, the APISIX frontend by its built-in `opentelemetry` plugin, all exporting OTLP traces to an OpenTelemetry Collector that delivers every span to the OAP both natively as OTLP and converted to Zipkin, so Grafana Tempo serves the same traces at `/otlp` and `/zipkin`. | Runs beside the SkyWa [...]
+| `otlp-trace`            | Deploy a second copy of the music application from its agentless images: the Java, Node.js and Python services instrumented at start by the OpenTelemetry auto-instrumentation, the APISIX frontend by its built-in `opentelemetry` plugin, every span exported straight to the OAP and stored natively, so Grafana Tempo serves this copy at `/otlp`. | Runs beside the SkyWalking-agent copy, in its own namespace on Kubernetes and its own network on Docker Compose, so bot [...]
 | `r3`                    | Deploy [R3](https://github.com/SkyAPM/R3) as RESTful URL recognition service.                                                                                                          |                                                                                                                                       |
 | `rocketmq-monitor`      | Deploy OpenTelemetry and export RocketMQ monitoring metrics to SkyWalking for analysis and display on UI.                                                                              |                                                                                                                                       |
 | `pulsar-monitor`        | Deploy OpenTelemetry and export Pulsar monitoring metrics to SkyWalking for analysis and display on UI.                                                                                |                                                                                                                                       |