feat(syslog): 优化大日志通道并增加异常连续消费兜底

- 128 KiB 及以上日志默认进入大日志 Topic - InfluxDB 原文及达梦大字段超过 256 KiB 时截断保存 - 截断内容追加 HB_XDR_TRUNCATED 标记 - InfluxDB 写入失败时继续执行规则解析和达梦写入 - 未知或系统性异常优先写入普通/大日志 DLT - DLT 写入失败时记录特殊告警并提交 offset，避免阻塞后续消息 - DLT 失败等待时间默认限制为 1 秒 - 增加连续性放行、DLT 成功及失败监控指标 - 更新 Docker Compose、v3 升级文档及相关测试
2026-09-12 19:22:52 +08:00
parent 3b6c9e4714
commit 0e9e6656db
64 changed files with 3189 additions and 63 deletions
@@ -0,0 +1,77 @@
+#!/usr/bin/env bash
+set -euo pipefail
+
+# 创建大日志主 Topic 及独立 Retry/DLT/Replay Topic。不删除消息、不修改消费组 offset。
+BOOTSTRAP="${BOOTSTRAP:-10.141.121.35:9092,10.141.121.50:9092,10.141.121.59:9092}"
+RAW_TOPIC="${RAW_TOPIC:-agent-01-syslog-topic}"
+LARGE_TOPIC="${LARGE_TOPIC:-${RAW_TOPIC}.large}"
+MAX_MESSAGE_BYTES="${MAX_MESSAGE_BYTES:-33554432}"
+RETENTION_MS="${RETENTION_MS:-604800000}"
+DLT_RETENTION_MS="${DLT_RETENTION_MS:-2592000000}"
+
+if [[ -z "${KAFKA_BIN:-}" ]]; then
+  for candidate in \
+    "$(cd -- "$(dirname -- "${BASH_SOURCE[0]}")" && pwd)" \
+    "/opt/kafka/kafka_2.13-3.6.1/bin" \
+    "/opt/kafka/bin" \
+    "/usr/local/kafka/bin"; do
+    if [[ -x "${candidate}/kafka-topics.sh" ]] && [[ -x "${candidate}/kafka-configs.sh" ]]; then
+      KAFKA_BIN="${candidate}"
+      break
+    fi
+  done
+fi
+if [[ -z "${KAFKA_BIN:-}" ]] && command -v kafka-topics.sh >/dev/null 2>&1; then
+  KAFKA_BIN=$(dirname -- "$(command -v kafka-topics.sh)")
+fi
+if [[ -z "${KAFKA_BIN:-}" ]]; then
+  echo "ERROR: 未找到 Kafka CLI；请设置 KAFKA_BIN" >&2
+  exit 1
+fi
+
+TOPICS="${KAFKA_BIN}/kafka-topics.sh"
+CONFIGS="${KAFKA_BIN}/kafka-configs.sh"
+COMMON=(--bootstrap-server "${BOOTSTRAP}")
+if [[ -n "${COMMAND_CONFIG:-}" ]]; then
+  [[ -r "${COMMAND_CONFIG}" ]] || { echo "ERROR: 无法读取 ${COMMAND_CONFIG}" >&2; exit 1; }
+  COMMON+=(--command-config "${COMMAND_CONFIG}")
+fi
+
+RAW_INFO=$("${TOPICS}" "${COMMON[@]}" --describe --topic "${RAW_TOPIC}")
+PARTITIONS=$(printf '%s\n' "${RAW_INFO}" | awk '
+  /PartitionCount:/ {for(i=1;i<=NF;i++) if($i=="PartitionCount:"){print $(i+1);exit}}
+')
+REPLICAS=$(printf '%s\n' "${RAW_INFO}" | awk '
+  /ReplicationFactor:/ {for(i=1;i<=NF;i++) if($i=="ReplicationFactor:"){print $(i+1);exit}}
+')
+if [[ ! "${PARTITIONS}" =~ ^[1-9][0-9]*$ ]] || [[ ! "${REPLICAS}" =~ ^[1-9][0-9]*$ ]]; then
+  echo "ERROR: 无法读取 ${RAW_TOPIC} 的分区数或副本数" >&2
+  exit 1
+fi
+
+declare -a TOPIC_NAMES=(
+  "${LARGE_TOPIC}"
+  "${LARGE_TOPIC}.RETRY.1M"
+  "${LARGE_TOPIC}.RETRY.10M"
+  "${LARGE_TOPIC}.RETRY.1H"
+  "${LARGE_TOPIC}.DLT"
+  "${LARGE_TOPIC}.REPLAY"
+)
+
+echo "Kafka=${BOOTSTRAP}，分区=${PARTITIONS}，副本=${REPLICAS}，单条上限=${MAX_MESSAGE_BYTES}"
+for topic in "${TOPIC_NAMES[@]}"; do
+  topic_retention="${RETENTION_MS}"
+  [[ "${topic}" == "${LARGE_TOPIC}.DLT" ]] && topic_retention="${DLT_RETENTION_MS}"
+  "${TOPICS}" "${COMMON[@]}" --create --if-not-exists \
+    --topic "${topic}" --partitions "${PARTITIONS}" --replication-factor "${REPLICAS}" \
+    --config cleanup.policy=delete \
+    --config "retention.ms=${topic_retention}" \
+    --config "max.message.bytes=${MAX_MESSAGE_BYTES}"
+  "${CONFIGS}" "${COMMON[@]}" --alter --entity-type topics --entity-name "${topic}" \
+    --add-config "cleanup.policy=delete,retention.ms=${topic_retention},max.message.bytes=${MAX_MESSAGE_BYTES}"
+  "${TOPICS}" "${COMMON[@]}" --describe --topic "${topic}"
+done
+
+echo "PASS: 大日志 Topic 已创建并统一配置。"
+echo "注意：Broker 的 replica.fetch.max.bytes 也必须不小于 ${MAX_MESSAGE_BYTES}。"
+echo "脚本未删除消息，也未修改任何消费组 offset。"