WIP: integrate OpenTelemetry SDK for metrics/logs export

.github/workflows/ci-tap.yml

@@ -96,11 +96,31 @@ jobs:
         run: |
           docker exec clickhouse-test clickhouse-client --multiquery < docker/init/00-schema.sql
 
+      - name: Start OTel Collector
+        run: |
+          export OTEL_DATA_DIR=/tmp/psch-otel-data
+          mkdir -p "$OTEL_DATA_DIR"
+          chmod 777 "$OTEL_DATA_DIR"
+          docker compose -f docker/docker-compose.otel.yml up -d
+          for i in {1..30}; do
+            if curl -sf http://localhost:13133/ >/dev/null 2>&1; then
+              echo "OTel Collector ready"
+              exit 0
+            fi
+            echo "Waiting for OTel Collector... ($i/30)"
+            sleep 1
+          done
+          echo "OTel Collector not ready after 30s"
+          docker logs psch-otel-collector
+          exit 1
+
       - name: Run TAP tests
         run: |
           export PATH="${{ env.PG_TAP_DIR }}/bin:$PATH"
           export PG_REGRESS="${{ env.PG_TAP_DIR }}/lib/postgresql/pgxs/src/test/regress/pg_regress"
           export PERL_LIB="${{ env.PG_TAP_DIR }}/lib/postgresql/pgxs/src/test/perl"
+          export OTEL_DATA_DIR=/tmp/psch-otel-data
+          export PROJECT_DIR="${{ github.workspace }}"
           prove -v --timer -I "$PERL_LIB" -I t t/*.pl
 
       - name: Upload logs on failure
@@ -113,3 +133,7 @@ jobs:
       - name: Cleanup ClickHouse
         if: always()
         run: docker rm -f clickhouse-test 2>/dev/null || true
+
+      - name: Cleanup OTel Collector
+        if: always()
+        run: docker compose -f docker/docker-compose.otel.yml down -v 2>/dev/null || true

.gitignore

@@ -5,6 +5,8 @@
 *.a
 .deps/
 build/
+CMakeCache.txt
+CMakeFiles/
 
 # Compile database
 compile_commands.json

.gitmodules

@@ -1,3 +1,6 @@
 [submodule "third_party/clickhouse-cpp"]
 	path = third_party/clickhouse-cpp
 	url = https://github.com/ClickHouse/clickhouse-cpp.git
+[submodule "third_party/opentelemetry-cpp"]
+	path = third_party/opentelemetry-cpp
+	url = https://github.com/open-telemetry/opentelemetry-cpp.git

CMakeLists.txt

@@ -25,12 +25,38 @@ option(WITH_OPENSSL "Enable OpenSSL for TLS connections" ON)
 include(PgStatChOpenSSL)
 pg_stat_ch_setup_openssl()
 
-# Add clickhouse-cpp library (statically linked, compiled with PIC for shared library linking)
+# Build all third-party libraries as static with PIC
 set(CMAKE_POSITION_INDEPENDENT_CODE ON)
-set(BUILD_SHARED_LIBS OFF CACHE BOOL "Build clickhouse-cpp as static library" FORCE)
+set(BUILD_SHARED_LIBS OFF CACHE BOOL "Build third-party libs as static" FORCE)
+
+# ---------------------------------------------------------------------------
+# Add opentelemetry-cpp FIRST (fetches gRPC + abseil via FetchContent).
+# This must come before clickhouse-cpp to avoid duplicate abseil targets.
+# ---------------------------------------------------------------------------
+include(FetchContent)
+set(WITH_OTLP_GRPC ON CACHE BOOL "Enable OTel OTLP gRPC exporter" FORCE)
+set(WITH_OTLP_HTTP OFF CACHE BOOL "" FORCE)
+set(BUILD_TESTING OFF CACHE BOOL "Disable tests" FORCE)
+set(WITH_BENCHMARK OFF CACHE BOOL "Disable benchmarks" FORCE)
+set(WITH_EXAMPLES OFF CACHE BOOL "Disable OTel examples" FORCE)
+set(WITH_FUNC_TESTS OFF CACHE BOOL "Disable OTel functional tests" FORCE)
+set(OPENTELEMETRY_INSTALL OFF CACHE BOOL "" FORCE)
+set(WITH_ABSEIL ON CACHE BOOL "Use Abseil for OTel" FORCE)
+# Always use vendored gRPC + abseil (never pick up system packages).
+# This prevents find_package(gRPC) from short-circuiting FetchContent
+# and leaving absl:: targets undefined.
+set(CMAKE_DISABLE_FIND_PACKAGE_gRPC TRUE)
+add_subdirectory(third_party/opentelemetry-cpp EXCLUDE_FROM_ALL)
+
+# ---------------------------------------------------------------------------
+# Add clickhouse-cpp AFTER opentelemetry-cpp.
+# Use WITH_SYSTEM_ABSEIL so it finds the abseil already built by gRPC above
+# (via our cmake/Findabsl.cmake shim).
+# ---------------------------------------------------------------------------
+set(WITH_SYSTEM_ABSEIL ON CACHE BOOL "Use abseil from gRPC build" FORCE)
 set(WITH_OPENSSL ${WITH_OPENSSL} CACHE BOOL "Pass OpenSSL setting to clickhouse-cpp" FORCE)
-set(BUILD_TESTS OFF CACHE BOOL "Disable clickhouse-cpp tests")
-set(BUILD_BENCHMARK OFF CACHE BOOL "Disable clickhouse-cpp benchmarks")
+set(BUILD_TESTS OFF CACHE BOOL "Disable clickhouse-cpp tests" FORCE)
+set(BUILD_BENCHMARK OFF CACHE BOOL "Disable clickhouse-cpp benchmarks" FORCE)
 add_subdirectory(third_party/clickhouse-cpp EXCLUDE_FROM_ALL)
 
 # Collect source files
@@ -50,10 +76,19 @@ target_include_directories(pg_stat_ch SYSTEM PRIVATE
   ${CMAKE_SOURCE_DIR}/third_party/clickhouse-cpp
   ${CMAKE_SOURCE_DIR}/third_party/clickhouse-cpp/contrib/absl
   ${CMAKE_SOURCE_DIR}/third_party/clickhouse-cpp/contrib/zstd
+  ${CMAKE_SOURCE_DIR}/third_party/opentelemetry-cpp/api/include
+  ${CMAKE_SOURCE_DIR}/third_party/opentelemetry-cpp/sdk/include
+  ${CMAKE_SOURCE_DIR}/third_party/opentelemetry-cpp/exporters/otlp/include
 )
 target_link_libraries(pg_stat_ch PRIVATE
   PostgreSQLServer::PostgreSQLServer
   clickhouse-cpp-lib
+  opentelemetry_api
+  opentelemetry_sdk
+  opentelemetry_exporter_otlp_grpc_metrics
+  opentelemetry_exporter_otlp_grpc_log
+  opentelemetry_metrics
+  opentelemetry_logs
 )
 if(WITH_OPENSSL)
   target_link_libraries(pg_stat_ch PRIVATE OpenSSL::SSL OpenSSL::Crypto)

cmake/Findabsl.cmake

@@ -0,0 +1,16 @@
+# Findabsl.cmake - Bridge find_package(absl) to vendored abseil targets.
+#
+# clickhouse-cpp calls find_package(absl REQUIRED) when WITH_SYSTEM_ABSEIL=ON.
+# The vendored gRPC (fetched by opentelemetry-cpp) builds abseil as a
+# subdirectory, so there's no abslConfig.cmake on disk. This Module-mode
+# find script checks that the required target exists and sets absl_FOUND.
+
+if(TARGET absl::int128)
+  set(absl_FOUND TRUE)
+else()
+  set(absl_FOUND FALSE)
+  if(absl_FIND_REQUIRED)
+    message(FATAL_ERROR "absl::int128 target not found. "
+      "Ensure opentelemetry-cpp (which fetches gRPC + abseil) is added before clickhouse-cpp.")
+  endif()
+endif()

docker/docker-compose.otel.yml

@@ -0,0 +1,10 @@
+services:
+  otel-collector:
+    image: otel/opentelemetry-collector-contrib:0.120.0
+    container_name: psch-otel-collector
+    ports:
+      - "4317:4317"
+      - "13133:13133"
+    volumes:
+      - ./otel-collector-config.yaml:/etc/otelcol-contrib/config.yaml:ro
+      - ${OTEL_DATA_DIR:-/tmp/psch-otel-data}:/otel-data

docker/otel-collector-config.yaml

@@ -0,0 +1,29 @@
+receivers:
+  otlp:
+    protocols:
+      grpc:
+        endpoint: 0.0.0.0:4317
+
+exporters:
+  file/logs:
+    path: /otel-data/logs.jsonl
+    format: json
+    flush_interval: 500ms
+  file/metrics:
+    path: /otel-data/metrics.jsonl
+    format: json
+    flush_interval: 500ms
+
+extensions:
+  health_check:
+    endpoint: 0.0.0.0:13133
+
+service:
+  extensions: [health_check]
+  pipelines:
+    logs:
+      receivers: [otlp]
+      exporters: [file/logs]
+    metrics:
+      receivers: [otlp]
+      exporters: [file/metrics]

include/config/guc.h

@@ -4,13 +4,16 @@
 #define PG_STAT_CH_GUC_H
 
 extern bool psch_enabled;
+extern bool psch_use_otel;
 extern char* psch_clickhouse_host;
 extern int psch_clickhouse_port;
 extern char* psch_clickhouse_user;
 extern char* psch_clickhouse_password;
 extern char* psch_clickhouse_database;
 extern bool psch_clickhouse_use_tls;
 extern bool psch_clickhouse_skip_tls_verify;
+extern char* psch_otel_endpoint;
+extern char* psch_hostname;
 extern int psch_queue_capacity;
 extern int psch_flush_interval_ms;
 extern int psch_batch_max;

scripts/run-tests.sh

@@ -17,7 +17,7 @@ usage() {
     echo "Usage: $0 <PG_VERSION|PG_PATH> [test_type] [test_filter]"
     echo "  PG_VERSION:  PostgreSQL version (16, 17, 18) - uses mise"
     echo "  PG_PATH:     Path to local PostgreSQL installation"
-    echo "  test_type:   regress, tap, isolation, stress, clickhouse, or all (default: all)"
+    echo "  test_type:   regress, tap, isolation, stress, clickhouse, otel, or all (default: all)"
     echo "  test_filter: (tap only) pattern to match test files, e.g., '021' for t/*021*.pl"
     echo ""
     echo "Examples:"
@@ -27,6 +27,7 @@ usage() {
     echo "  $0 ../postgres/install_tap tap 021  # Run only t/*021*.pl test"
     echo "  $0 18 stress                    # Run only stress test against PG 18"
     echo "  $0 18 clickhouse                # Run ClickHouse integration tests (requires Docker)"
+    echo "  $0 18 otel                      # Run OTel integration tests (requires Docker)"
     exit 1
 }
 
@@ -250,6 +251,68 @@ run_clickhouse() {
         t/010_clickhouse_export.pl t/011_clickhouse_reconnect.pl
 }
 
+# Run OTel integration tests
+run_otel() {
+    log_info "Running OTel integration tests..."
+
+    # Check if Docker is available
+    if ! command -v docker &> /dev/null; then
+        log_error "Docker is required for OTel tests"
+        return 1
+    fi
+
+    # Check if OTel Collector is running, start if not
+    if ! curl -sf http://localhost:13133/ >/dev/null 2>&1; then
+        log_info "Starting OTel Collector container..."
+
+        export OTEL_DATA_DIR="${OTEL_DATA_DIR:-/tmp/psch-otel-data}"
+        mkdir -p "${OTEL_DATA_DIR}"
+        chmod 777 "${OTEL_DATA_DIR}"
+
+        docker compose -f docker/docker-compose.otel.yml up -d
+        # Wait for health check
+        for i in $(seq 1 30); do
+            if curl -sf http://localhost:13133/ >/dev/null 2>&1; then
+                break
+            fi
+            sleep 1
+        done
+
+        if ! curl -sf http://localhost:13133/ >/dev/null 2>&1; then
+            log_error "OTel Collector failed to start"
+            return 1
+        fi
+    fi
+
+    local perl_lib="${PG_LIB}/pgxs/src/test/perl"
+    if [[ ! -d "${perl_lib}" ]]; then
+        log_warn "PostgreSQL TAP test modules not found at ${perl_lib}"
+        log_warn "OTel tests require PostgreSQL built with --enable-tap-tests"
+        log_warn "Skipping OTel tests."
+        return 0
+    fi
+
+    local pg_regress="${PG_LIB}/pgxs/src/test/regress/pg_regress"
+    if [[ ! -x "${pg_regress}" ]]; then
+        log_warn "pg_regress not found at ${pg_regress}"
+        log_warn "Skipping OTel tests."
+        return 0
+    fi
+
+    # Clean up stale test data directories
+    rm -rf tmp_check
+
+    # Set env vars for the test helper module
+    export PROJECT_DIR="${PROJECT_DIR}"
+    export OTEL_DATA_DIR="${OTEL_DATA_DIR:-/tmp/psch-otel-data}"
+
+    # Run only OTel-related tests
+    PG_REGRESS="${pg_regress}" prove -v --timer \
+        -I "${perl_lib}" \
+        -I t \
+        t/024_otel_export.pl t/025_otel_reconnect.pl
+}
+
 # Main execution
 case "${TEST_TYPE}" in
     regress)
@@ -267,6 +330,9 @@ case "${TEST_TYPE}" in
     clickhouse)
         run_clickhouse
         ;;
+    otel)
+        run_otel
+        ;;
     all)
         run_regress
         run_tap

src/config/guc.cc

@@ -12,13 +12,16 @@ extern "C" {
 
 // GUC variable storage
 bool psch_enabled = true;
+bool psch_use_otel = false;
 char* psch_clickhouse_host = nullptr;
 int psch_clickhouse_port = 9000;
 char* psch_clickhouse_user = nullptr;
 char* psch_clickhouse_password = nullptr;
 char* psch_clickhouse_database = nullptr;
 bool psch_clickhouse_use_tls = false;
 bool psch_clickhouse_skip_tls_verify = false;
+char* psch_otel_endpoint = nullptr;
+char* psch_hostname = nullptr;
 int psch_queue_capacity = 131072;
 int psch_flush_interval_ms = 200;
 int psch_batch_max = 200000;
@@ -79,6 +82,16 @@ void PschInitGuc(void) {
       0,                                                      // flags
       nullptr, nullptr, nullptr);                             // hooks
 
+  DefineCustomBoolVariable(
+      "pg_stat_ch.use_otel",
+      "Send metrics through OpenTelemetry instead of ClickHouse.",
+      "When enabled, stats will be sent to an OTel endpoint instead of ClickHouse.",
+      &psch_use_otel,
+      false,
+      PGC_POSTMASTER,
+      0,
+      nullptr, nullptr, nullptr);
+
   DefineCustomStringVariable(
       "pg_stat_ch.clickhouse_host",
       "ClickHouse server hostname.",
@@ -150,6 +163,26 @@ void PschInitGuc(void) {
       0,
       nullptr, nullptr, nullptr);
 
+  DefineCustomStringVariable(
+      "pg_stat_ch.otel_endpoint",
+      "OpenTelemetry gRPC endpoint (host:port).",
+      nullptr,
+      &psch_otel_endpoint,
+      "localhost:4317",
+      PGC_POSTMASTER,
+      0,
+      nullptr, nullptr, nullptr);
+
+  DefineCustomStringVariable(
+      "pg_stat_ch.hostname",
+      "The hostname of the current machine.",
+      nullptr,
+      &psch_hostname,
+      "",
+      PGC_POSTMASTER,
+      0,
+      nullptr, nullptr, nullptr);
+
   DefineCustomIntVariable(
       "pg_stat_ch.queue_capacity",
       "Maximum number of events in the shared memory queue (must be a power of 2).",

src/config/guc.h

@@ -10,11 +10,14 @@ extern "C" {
 
 // GUC variables (defined in guc.cc)
 extern bool psch_enabled;
+extern bool psch_use_otel;
 extern char* psch_clickhouse_host;
 extern int psch_clickhouse_port;
 extern char* psch_clickhouse_user;
 extern char* psch_clickhouse_password;
 extern char* psch_clickhouse_database;
+extern char* psch_otel_endpoint;
+extern char* psch_hostname;
 extern int psch_queue_capacity;
 extern int psch_flush_interval_ms;
 extern int psch_batch_max;