feat: introduce distributed epoch-level tracing

Cargo.toml

@@ -106,8 +106,6 @@ opt-level = 3
 opt-level = 3
 [profile.ci-dev.package."indextree"]
 opt-level = 3
-[profile.ci-dev.package."task_stats_alloc"]
-opt-level = 3
 
 # The profile used for deterministic simulation tests in CI.
 # The simulator can only run single-threaded, so optimization is required to make the running time

ci/workflows/pull-request.yml

@@ -79,7 +79,7 @@ steps:
           config: ci/docker-compose.yml
           mount-buildkite-agent: true
       - ./ci/plugins/upload-failure-logs
-    timeout_in_minutes: 12
+    timeout_in_minutes: 15
 
   - label: "end-to-end test for opendal (parallel)"
     command: "ci/scripts/e2e-test-parallel-for-opendal.sh -p ci-dev"

proto/stream_plan.proto

@@ -93,7 +93,7 @@ message Barrier {
     ResumeMutation resume = 8;
   }
   // Used for tracing.
-  bytes span = 2;
+  map<string, string> tracing_context = 2;
   // Whether this barrier do checkpoint
   bool checkpoint = 9;
 

proto/stream_service.proto

@@ -68,6 +68,7 @@ message InjectBarrierResponse {
 message BarrierCompleteRequest {
   string request_id = 1;
   uint64 prev_epoch = 2;
+  map<string, string> tracing_context = 3;
 }
 message BarrierCompleteResponse {
   message CreateMviewProgress {

risedev.yml

@@ -24,14 +24,18 @@ profile:
 
       # If you want to use aws-s3, configure AK and SK in env var and enable the following lines:
       # - use: aws-s3
-        # bucket: test-bucket
+      #   bucket: test-bucket
 
       # If you want to create CDC source table, uncomment the following line
       # - use: connector-node
 
       # if you want to enable etcd backend, uncomment the following lines.
       # - use: etcd
       #   unsafe-no-fsync: true
+
+      # If you want to enable tracing, uncomment the following line.
+      # - use: jaeger
+
       - use: meta-node
       - use: compute-node
       - use: frontend
@@ -43,9 +47,6 @@ profile:
       # - use: prometheus
       # - use: grafana
 
-      # If you want to enable tracing, uncomment the following line.
-      # - use: jaeger
-
       # If you want to create source from Kafka, uncomment the following lines
       # Note that kafka depends on zookeeper, so zookeeper must be started beforehand.
       # - use: zookeeper
@@ -898,6 +899,9 @@ template:
     # AWS s3 bucket used by the cluster
     provide-aws-s3: "aws-s3*"
 
+    # Jaeger used by this meta node
+    provide-jaeger: "jaeger*"
+
     # Whether to enable in-memory pure KV state backend
     enable-in-memory-kv-state-backend: false
 
@@ -969,6 +973,9 @@ template:
     # Meta-nodes used by this frontend instance
     provide-meta-node: "meta-node*"
 
+    # Jaeger used by this frontend instance
+    provide-jaeger: "jaeger*"
+
     # If `user-managed` is true, this service will be started by user with the above config
     user-managed: false
 
@@ -991,9 +998,12 @@ template:
     # Minio instances used by this compactor
     provide-minio: "minio*"
 
-    # Meta-nodes used by this compute node
+    # Meta-nodes used by this compactor
     provide-meta-node: "meta-node*"
 
+    # Jaeger used by this compator
+    provide-jaeger: "jaeger*"
+
     # If `user-managed` is true, this service will be started by user with the above config
     user-managed: false
 
@@ -1039,6 +1049,12 @@ template:
     # Dashboard listen port of Jaeger
     dashboard-port: 16680
 
+    # gRPC listen address of the OTLP collector
+    otlp-address: "127.0.0.1"
+
+    # gRPC listen port of the OTLP collector
+    otlp-port: 4317
+
     # Jaeger has a lot of ports open, and we don't want to make this config more complex.
     # So we keep the default value of jaeger instead of making it part of RiseDev config.
 
@@ -1064,7 +1080,6 @@ template:
 
     # access key, secret key and region should be set in aws config (either by env var or .aws/config)
 
-
   # Apache Kafka service
   kafka:
     # Id to be picked-up by services