Patroni集群中vip-manager无法启用虚拟IP的问题求助

安装vip-manager用于管理PostgreSQL Patroni集群的可迁移虚拟IP，但vip-manager无法启用虚拟IP地址。

环境信息 #

• RHEL 9.3
• etcd版本：3.4.31
• patroni版本：3.0.4
• vip-manager版本：2.3.0

vip-manager配置 #

interval: 1000
trigger-key: "/pg_cluster/postgres/leader"
trigger-value: "sp-smaxpsql1"
ip: 10.164.20.180 # the virtual ip address to manage
netmask: 24 # netmask for the virtual ip
interface: ens33 #interface to which the virtual ip will be added
hosting-type: basic # possible values: basic, or hetzner.
dcs-type: etcd # etcd or consul
dcs-endpoints:
  - http://10.164.20.177:2379
  - http://10.164.20.178:2379
  - http://10.164.20.179:2379
retry-num: 2
retry-after: 250  #in milliseconds
verbose: true

etcd配置 #

ETCD_LISTEN_PEER_URLS="http://10.164.20.177:2380"
ETCD_LISTEN_CLIENT_URLS="http://10.164.20.177:2379,http://localhost:2379"
ETCD_NAME="sp-smaxpsql1"
[Clustering]
ETCD_INITIAL_ADVERTISE_PEER_URLS="http://10.164.20.177:2380"
ETCD_ADVERTISE_CLIENT_URLS="http://10.164.20.177:2379"
ETCD_INITIAL_CLUSTER="sp-smaxpsql1=http://10.164.20.177:2380,sp-smaxpsql2=http://10.164.20.178:2380,sp-smaxpsql3=http://10.164.20.179:2380"
ETCD_INITIAL_CLUSTER_TOKEN="postgresql-etcd-cluster"
ETCD_INITIAL_CLUSTER_STATE="new"
ETCD_DATA_DIR="/var/lib/etcd"

Patroni配置 #

scope: postgres
namespace: /pg_cluster/
name: sp-smaxpsql1
restapi:
  listen: sp-smaxpsql1:8008
  connect_address: sp-smaxpsql1:8008
etcd:
  hosts: 10.164.20.177:2379, 10.164.20.178:2379, 10.164.20.179:2379
bootstrap:
  dcs:
    ttl: 30
    loop_wait: 10
    retry_timeout: 10
    maximum_lag_on_failover: 1048576
    maximum_lag_on_syncnode: 15000000
    synchronous_mode: false
    postgresql:
      use_pg_rewind: true
      use_slots: true
      parameters:
        shared_buffers: 8GB
        work_mem: 13981kB
        maintenance_work_mem: 2GB
        max_worker_processes: 16
        wal_buffers: 64MB
        max_wal_size: 2GB
        min_wal_size: 8GB
        effective_cache_size: 24GB
        fsync: on
        checkpoint_completion_target: 0.9
        log_rotation_size: 100MB
        listen_addresses: "*"
        max_connections: 2000
        temp_buffers: 4MB
        default_statistics_target: 100
        random_page_cost: 1.1
        effective_io_concurrency: 200
        ssl: true
        ssl_cert_file: /var/lib/pgsql/12/data/server.crt
        ssl_key_file: /var/lib/pgsql/12/data/server.key
  initdb:
    - encoding: UTF8
    - data-checksums
  pg_hba:
    - host replication replicator 127.0.0.1/32 md5
    - host replication replicator 10.164.20.177/32 md5
    - host replication replicator 10.164.20.178/32 md5
    - host replication replicator 10.164.20.179/32 md5
    - host all all 0.0.0.0/0 md5
users:
  admin:
    password: admin
    options:
      - createrole
      - createdb

postgresql:
  listen: sp-smaxpsql1:5432
  connect_address: sp-smaxpsql1:5432
  data_dir: /var/lib/pgsql/12/data
  bin_dir: /usr/pgsql-12/bin
  pgpass: /tmp/pgpass
  authentication:
    replication:
      username: replicator
      password: <*********>
    superuser:
      username: postgres
      password: <*********>
    rewind:
      username: pgrewind
      password: <*********>
tags:
  nofailover: false
  noloadbalance: false
  clonefrom: false
  nosync: true

主节点vip-manager日志 #

Mar 27 15:24:26 sp-smaxpsql1 vip-manager[68389]: 2024/03/27 15:24:26 IP address 10.164.20.180/24 state is false, desired false
Mar 27 15:24:36 sp-smaxpsql1 vip-manager[68389]: 2024/03/27 15:24:36 IP address 10.164.20.180/24 state is false, desired false
Mar 27 15:24:46 sp-smaxpsql1 vip-manager[68389]: 2024/03/27 15:24:46 IP address 10.164.20.180/24 state is false, desired false
Mar 27 15:24:56 sp-smaxpsql1 vip-manager[68389]: 2024/03/27 15:24:56 IP address 10.164.20.180/24 state is false, desired false
Mar 27 15:25:06 sp-smaxpsql1 vip-manager[68389]: 2024/03/27 15:25:06 IP address 10.164.20.180/24 state is false, desired false
Mar 27 15:25:16 sp-smaxpsql1 vip-manager[68389]: 2024/03/27 15:25:16 IP address 10.164.20.180/24 state is false, desired false
Mar 27 15:25:26 sp-smaxpsql1 vip-manager[68389]: 2024/03/27 15:25:26 IP address 10.164.20.180/24 state is false, desired false
Mar 27 15:25:36 sp-smaxpsql1 vip-manager[68389]: 2024/03/27 15:25:36 IP address 10.164.20.180/24 state is false, desired false
Mar 27 15:25:46 sp-smaxpsql1 vip-manager[68389]: 2024/03/27 15:25:46 IP address 10.164.20.180/24 state is false, desired false

解决方案 #

从日志看，vip-manager认为当前节点不需要启用VIP（desired false），核心原因是etcd中trigger-key对应的值和trigger-value不匹配，或者vip-manager无法正确读取etcd中的键值。按以下步骤排查：

1. 验证etcd中的leader键值
   在任意节点执行etcdctl命令查看/pg_cluster/postgres/leader的值：

ETCDCTL_API=3 etcdctl --endpoints=http://10.164.20.177:2379,http://10.164.20.178:2379,http://10.164.20.179:2379 get /pg_cluster/postgres/leader

如果返回的值不是sp-smaxpsql1，说明当前主节点不是sp-smaxpsql1，或者Patroni没有正确更新etcd中的leader键。先确认Patroni集群的主节点状态：

patronictl -c /path/to/patroni.yml list

确保sp-smaxpsql1显示为Leader。

2. 修正vip-manager的trigger配置

• 如果etcd中leader键的值是节点的IP而非主机名，把vip-manager的trigger-value改成对应IP（比如10.164.20.177）
• 检查trigger-key路径是否和Patroni配置一致：Patroni的namespace是/pg_cluster/，scope是postgres，所以leader键的正确路径应该是/pg_cluster/postgres/leader，当前配置是对的，但要确认etcd中确实存在这个键。

3. 检查vip-manager的权限
   vip-manager需要root权限才能添加/删除虚拟IP，确保服务是以root身份运行：

ps aux | grep vip-manager

如果不是root，修改systemd服务配置（通常在/etc/systemd/system/vip-manager.service），添加User=root，然后重启服务：

systemctl daemon-reload
systemctl restart vip-manager

4. 手动测试添加VIP
   在主节点手动执行添加VIP的命令，验证网络接口和IP是否可用：

ip addr add 10.164.20.180/24 dev ens33

执行后用ip addr show ens33检查是否添加成功，如果失败，可能是IP已被占用，或者网络接口名称错误（比如实际接口是ens3而非ens33，用ip link确认）。

5. 验证vip-manager与etcd的连接
   在vip-manager所在节点测试etcd连接：

ETCDCTL_API=3 etcdctl --endpoints=http://10.164.20.177:2379 endpoint health

如果连接失败，检查etcd服务是否运行，防火墙是否开放2379端口，或者etcd配置是否限制了客户端连接。

6. 重启vip-manager服务
   修改配置后，重启vip-manager并查看日志：

systemctl restart vip-manager
journalctl -u vip-manager -f

观察日志中是否出现desired true的记录，此时VIP应该会被自动添加。

内容的提问来源于stack exchange，提问作者Murat Suluhan