Django应用服务器间歇性Pooler Error问题排查求助

HTTP服务器间歇性触发连接池错误：服务器连接崩溃？错误，已排查并修复系统中的长事务，但问题仍未解决。该错误随机出现，与流量负载无关联。

pgbouncer错误日志 #

2025-01-31 11:41:07.067 1 日志 C-0x55d7847ac110: logs/server_application_api@127.0.0.1:34186 关闭原因：客户端发起关闭请求（存活时长=0）
2025-01-31 11:41:07.051 1 日志 C-0x55d7847ac110: logs/server_application_api@127.0.0.1:34186 登录尝试：数据库=test 用户=server_application_api TLS=未启用
2025-01-31 11:41:06.988 1 日志 C-0x55d7847ac110: logs/server_application_api@127.0.0.1:34096 关闭原因：客户端发起关闭请求（存活时长=0）
2025-01-31 11:41:06.968 1 日志 C-0x55d7847ae820: logs/server_application_api@127.0.0.1:34048 关闭原因：客户端发起关闭请求（存活时长=0）
2025-01-31 11:41:06.952 1 日志 C-0x55d7847ac110: logs/server_application_api@127.0.0.1:34096 登录尝试：数据库=test 用户=server_application_api TLS=未启用
2025-01-31 11:41:06.948 1 警告 C-0x55d7847ac110: logs/server_application_api@127.0.0.1:40674 连接池错误：服务器连接崩溃？

CloudWatch指标 #

CloudWatch指标图显示数据库连接相关指标存在异常波动

当前pgbouncer配置 #

pgbouncer.ini.ctmpl: |
    [databases]
    logs =  dbname='sample_db'    host='sample_host'   port=5432   user=sample_user_name   pool_size=32
    [pgbouncer]
    listen_addr = 0.0.0.0
    listen_port = 6432
    unix_socket_dir = /var/run/postgresql
    auth_type = plain
    auth_file = /etc/pgbouncer/userlist.txt
    admin_users = sample_user_name
    stats_users = stats, root
    pool_mode = session
    server_reset_query = DISCARD ALL
    ignore_startup_parameters = extra_float_digits
    server_check_query = select 1
    server_check_delay = 30
    max_client_conn = 100
    default_pool_size = 20
    reserve_pool_size = 5
    reserve_pool_timeout = 3
    max_db_connections = 50
    max_user_connections = 50
    log_connections = 1
    log_disconnections = 1
    log_pooler_errors = 1
    server_round_robin = 0
    server_lifetime = 1200
    server_idle_timeout = 60
    server_connect_timeout = 15
    server_login_retry = 15
    query_timeout = 0
    query_wait_timeout = 0
    client_idle_timeout = 0
    client_login_timeout = 60
    autodb_idle_timeout = 3600
    pkt_buf = 2048
    listen_backlog = 128
    tcp_defer_accept = 0
    tcp_socket_buffer = 0
    tcp_keepalive = 1
    tcp_keepcnt = 0
    tcp_keepidle = 0
    tcp_keepintvl = 0
    dns_max_ttl = 15
    dns_zone_check_period = 0
    pidfile = /var/log/pgbouncer/pgbouncer.pid
    idle_transaction_timeout = 15

（注：配置项关键字保留英文，为行业通用术语）

问题更新 #

已解决所有慢事务，消除了空闲查询，但仍出现突发的499请求峰值，说明空闲事务是错误的结果而非原因。

Nginx错误日志 #

{"connection": "13", "connection_requests": "1", "pid": "7", "request_id": "06ee9ab04bf", "request_length": "139", "remote_addr": "172.31.189.8", "remote_user": "", "remote_port": "2118", "time": "2025-08-06T14:03:40+00:00", "request": "GET /health_check HTTP/1.1", "full_path": "/health_check", "args": "", "downstream_service": "", "status": "499", "body_bytes_sent": "0", "bytes_sent": "0", "http_referer": "", "http_user_agent": "ELB-HealthChecker/2.0", "http_x_forwarded_for": "", "http_host": "172.24.93.01", "server_name": "", "request_time": "2.001", "upstream": "0.0.0.0:8000", "upstream_connect_time": "-", "upstream_header_time": "-", "upstream_response_time": "-", "upstream_response_length": "0", "upstream_cache_status": "", "method": "GET", "http_cf_ray": "" }

{"connection": "13", "connection_requests": "6", "pid": "7", "request_id": "d8eb51bad5a99", "request_length": "1346", "remote_addr": "172.31.22.21", "remote_user": "", "remote_port": "2944", "time": "2025-08-06T14:03:40+00:00", "request": "POST /api/management/callback/1292 HTTP/1.1", "full_path": "/api/management/callback/1292", "args": "", "downstream_service": "", "status": "499", "body_bytes_sent": "0", "bytes_sent": "0", "http_referer": "", "http_user_agent": "", "http_x_forwarded_for": "172.31.173.23, 172.31.253.13, 172.22.172.11", "http_host": "logs-server.prod-sgp-k8s.xyz.io", "server_name": "", "request_time": "5.003", "upstream": "0.0.0.0:8000", "upstream_connect_time": "-", "upstream_header_time": "-", "upstream_response_time": "-", "upstream_response_length": "0", "upstream_cache_status": "", "method": "POST", "http_cf_ray": "" }

（注：Nginx日志字段保留英文，为行业通用格式）

内容的提问来源于stack exchange，提问作者Dev