Merge pull request #213 from kyuengmanKim/refactor/code-updates

Refactor/log code updates

Author: kyuengmanKim

java-module/ansible/playbooks/install-agent/roles/agent/agent-install-files/cmp-agent/fluent-bit/etc/fluent-bit.conf

@@ -54,5 +54,5 @@
     name loki
     host ${LOKI_HOST}
     match syslog
-    labels host=${HOSTNAME}, uuid=${UUID}, credential_id=${credential_id}, cloud_service=${cloud_service}, type=${type}
+    labels host=${HOSTNAME}, NS_ID=${NS_ID}, MCI_ID=${MCI_ID}, VM_ID=${VM_ID}
     label_keys $service, $level, $source

java-module/docker-compose.yaml

@@ -6,6 +6,7 @@ networks:
 
 volumes:
   grafana_shared_config:
+  shared_logs:
 
 services:
   mc-observability-manager:
@@ -82,6 +83,7 @@ services:
     volumes:
       - ./mc-o11y-manager/config:/config
       - ./mc-o11y-manager/applog:/applog
+      - shared_logs:/applog/mc-o11y-manager:rw
       - grafana_shared_config:/grafana_config:ro
       # - ./manager-conf:/etc/mc-observability-agent/conf:rw
     healthcheck:
@@ -549,6 +551,7 @@ services:
       - OTEL_EXPORTER_OTLP_TRACES_ENDPOINT=${OTEL_EXPORTER_OTLP_TRACES_ENDPOINT:-http://mc-observability-tempo:4318/v1/traces}
     volumes:
       - ../python/insight/log:/mc-insight/log
+      - shared_logs:/mc-insight/log:rw
     networks:
       - internal_network
       - external_network
@@ -767,3 +770,33 @@ services:
     init: true
     networks:
       - internal_network
+
+  mc-observability-log-collector:
+    image: fluent/fluent-bit:3.2.4
+    container_name: mc-observability-log-collector
+    restart: always
+    networks:
+      - internal_network
+      - external_network
+    depends_on:
+      mc-observability-loki:
+        condition: service_healthy
+    environment:
+      - TZ=Asia/Seoul
+      - LOKI_HOST=mc-observability-loki
+      - LOKI_PORT=3100
+    volumes:
+      - shared_logs:/logs:ro
+      - ./log-conf/fluent-bit.conf:/fluent-bit/etc/fluent-bit.conf:ro
+      - ./log-conf/parsers.conf:/fluent-bit/etc/parsers.conf:ro
+      - ./log-conf/extract_component.lua:/fluent-bit/etc/extract_component.lua:ro
+    command:
+      - /fluent-bit/bin/fluent-bit
+      - -c
+      - /fluent-bit/etc/fluent-bit.conf
+    healthcheck:
+      test: ["CMD-SHELL", "wget --quiet --spider http://127.0.0.1:2020/api/v1/health || exit 1"]
+      interval: 10s
+      timeout: 5s
+      retries: 3
+      start_period: 10s

java-module/log-conf/extract_component.lua

@@ -0,0 +1,6 @@
+function extract_component(tag, timestamp, record)
+         if record["attributes"] and record["attributes"]["component"] then
+             record["component"] = record["attributes"]["component"]
+         end
+         return 1, timestamp, record
+     end

java-module/log-conf/fluent-bit.conf

@@ -0,0 +1,45 @@
+[SERVICE]
+    flush        5
+    daemon       Off
+    log_level    info
+    parsers_file /fluent-bit/etc/parsers.conf
+    HTTP_Server  On
+    HTTP_Listen  0.0.0.0
+    HTTP_PORT    2020
+
+[INPUT]
+    name              tail
+    tag               mc-observability-manager.logs
+    path              /logs/mc-o11y-manager.log
+    parser            json
+    read_from_head    false
+    refresh_interval  5
+
+[INPUT]
+    name              tail
+    tag               mc-observability-insight.logs
+    path              /logs/mc-observability-insight.log
+    parser            json
+    read_from_head    false
+    refresh_interval  5
+
+[FILTER]
+    Name        modify
+    Match_Regex ^mc-observability-.*\.logs$
+    Add         system  mc-observability
+
+[FILTER]
+    Name    lua
+    Match   *
+    script  /fluent-bit/etc/extract_component.lua
+    call    extract_component
+
+[OUTPUT]
+    name   loki
+    match  *
+    host   ${LOKI_HOST}
+    port   ${LOKI_PORT}
+    label_keys $system, $severity_text, $component
+    line_format json
+    auto_kubernetes_labels off
+    remove_keys component

java-module/log-conf/parsers.conf

@@ -0,0 +1,6 @@
+[PARSER]
+    Name   json
+    Format json
+    Time_Key timestamp
+    Time_Format %Y-%m-%dT%H:%M:%S.%LZ
+    Time_Keep On

java-module/mc-o11y-manager/src/main/resources/logback-spring.xml

@@ -22,7 +22,7 @@
         <file>${LOG_PATH}/${APPLICATION_NAME}/${LOG_FILE_NAME}.log</file>
         <encoder class="ch.qos.logback.classic.encoder.PatternLayoutEncoder">
             <pattern>
-                {"timestamp": "%d{yyyy-MM-dd'T'HH:mm:ss.SSS'Z'}","trace_id": "%X{trace_id}","span_id": "%X{span_id}","trace_flags": "%X{trace_flags}","severity_text": "%level", "attributes": {"component": "%X{component}","code.function": "%C.%M","code.file": "%F", "code.lineno": "%L"},"body": "%message"} \n
+                {"timestamp": "%d{yyyy-MM-dd'T'HH:mm:ss.SSS'Z'}","trace_id": "%X{trace_id}","span_id": "%X{span_id}","trace_flags": "%X{trace_flags}","severity_text": "%level", "attributes": {"component": "mc-observability-manager","code.function": "%C.%M","code.file": "%F", "code.lineno": "%L"},"body": "%message"} \n
             </pattern>
         </encoder>
         <rollingPolicy class="ch.qos.logback.core.rolling.SizeAndTimeBasedRollingPolicy">

java-module/scripts/init.sh

@@ -35,10 +35,7 @@ async def astart(self):
             return self.session
 
         except Exception as e:
-            logger.error(f"Failed to start Grafana MCP context: {e}")
-            import traceback
-
-            traceback.print_exc()
+            logger.exception(f"Failed to start Grafana MCP context: {e}")
             await self.astop()
             raise
 

java-module/scripts/install-manager.sh

@@ -35,10 +35,7 @@ async def astart(self):
             return self.session
 
         except Exception as e:
-            logger.error(f"Failed to start InfluxDB MCP context: {e}")
-            import traceback
-
-            traceback.print_exc()
+            logger.exception(f"Failed to start InfluxDB MCP context: {e}")
             await self.astop()
             raise