主/从模式下的 pgpool-ii：我怎样才能最容易地触发故障转移？

Question

所以我正在使用一些本地虚拟机测试一些玩具 postgresql 基础架构，以确定 pgpool 在故障转移时的行为。我已经配置了一个基本设置，其中有两台数据库机器（192.168.0.2 和 192.168.0.3）和一台 pgpool 机器（192.168.0.4）。 192.168.0.3 已使用流复制设置为 192.168.0.2 的从属设备。 pgpool-ii 已使用以下配置：

listen_addresses = '*'
backend_hostname0 = '192.168.0.2'
backend_port0 = 5432
backend_weight0 = 1
backend_data_directory0 = '/var/lib/postgresql/9.4/main/'
backend_flag0 = 'ALLOW_TO_FAILOVER'
backend_hostname1 = '192.168.0.3'
backend_port1 = 5432
backend_weight1 = 1
backend_data_directory1 = '/var/lib/postgresql/9.4/main/'
backend_flag1 = 'ALLOW_TO_FAILOVER'
enable_pool_hba = on
replication_mode = false
master_slave_mode = on
master_slave_sub_mode = 'stream'
fail_over_on_backend_error = true
failover_command = '/root/pgpool_failover_stream.sh %d %H /tmp/postgresql.trigger.5432'
load_balance_mode = false

我已经确认这一切正常。也就是说，当我更改 master 数据库时，复制正在工作，我可以通过示例应用程序连接到 master、slave 和 pgpool-ii 并获得我期望的结果。

现在，我启动了一个连接到 pgpool 的长时间运行的应用程序，然后尝试通过 SSH 连接到主数据库服务器并强制结束 postgres 任务（service postgresql stop 作为 root）来进行故障转移。我的应用程序继续正确执行查询，但没有发生故障转移（脚本尚未运行）。我什至测试过直接连接到主数据库，当我停止 postgres 服务时，我确实最终导致应用程序崩溃。

我做错了吗？我没有正确配置我的 pgpool 吗？还是有更好的方法来触发故障转移？

编辑：根据要求，这里是第一个错误发生的日志部分：

...
2016-03-15 18:47:15: pid 1232: DEBUG:  initializing backend status
2016-03-15 18:47:15: pid 1231: DEBUG:  initializing backend status
2016-03-15 18:47:15: pid 1230: DEBUG:  initializing backend status
2016-03-15 18:47:15: pid 1209: ERROR:  failed to authenticate
2016-03-15 18:47:15: pid 1209: DETAIL:  invalid authentication message response type, Expecting 'R' and received 'E'
2016-03-15 18:47:15: pid 1209: LOG:  find_primary_node: checking backend no 1
2016-03-15 18:47:15: pid 1209: ERROR:  failed to authenticate
2016-03-15 18:47:15: pid 1209: DETAIL:  invalid authentication message response type, Expecting 'R' and received 'E'
2016-03-15 18:47:15: pid 1209: DEBUG:  find_primary_node: no primary node found
...

奇怪的是，我仍然可以连接到 pgpool 并执行查询，所以很明显我不明白那里的东西。

编辑 2：这些是我在主服务器上 service postgresql shutdown 之后得到的错误。我展示了一切，直到 pgpool 开始关闭。

...
2016-03-16 17:24:57: pid 1012: DEBUG:  session context: clearing doing extended query messaging. DONE
2016-03-16 17:24:57: pid 1012: DEBUG:  session context: setting doing extended query messaging. DONE
2016-03-16 17:24:57: pid 1012: DEBUG:  session context: setting query in progress. DONE
2016-03-16 17:24:57: pid 1012: DEBUG:  reading backend data packet kind
2016-03-16 17:24:57: pid 1012: DETAIL:  backend:0 of 2 kind = 'E'
2016-03-16 17:24:57: pid 1012: DEBUG:  processing backend response
2016-03-16 17:24:57: pid 1012: DETAIL:  received kind 'E'(45) from backend
2016-03-16 17:24:57: pid 1012: ERROR:  unable to forward message to frontend
2016-03-16 17:24:57: pid 1012: DETAIL:  FATAL error occured on backend
2016-03-16 17:24:57: pid 1012: DEBUG:  session context: setting query in progress. DONE
2016-03-16 17:24:57: pid 1012: DEBUG:  decide where to send the queries
2016-03-16 17:24:57: pid 1012: DETAIL:  destination = 3 for query= "DISCARD ALL"
2016-03-16 17:24:57: pid 1012: DEBUG:  waiting for query response
2016-03-16 17:24:57: pid 1012: DETAIL:  waiting for backend:0 to complete the query
2016-03-16 17:24:57: pid 1012: FATAL:  unable to read data from DB node 0
2016-03-16 17:24:57: pid 1012: DETAIL:  EOF encountered with backend
2016-03-16 17:24:57: pid 998: DEBUG:  reaper handler
2016-03-16 17:24:57: pid 998: LOG:  child process with pid: 1012 exits with status 256
2016-03-16 17:24:57: pid 998: LOG:  fork a new child process with pid: 1033
2016-03-16 17:24:57: pid 998: DEBUG:  reaper handler: exiting normally
2016-03-16 17:24:57: pid 1033: DEBUG:  initializing backend status
2016-03-16 17:25:02: pid 1031: DEBUG:  PCP child receives shutdown request signal 2
2016-03-16 17:25:02: pid 1029: LOG:  child process received shutdown request signal 2
...

请注意，当主服务器关闭时，我的示例应用程序实际上确实死了。

编辑 3：在正确设置 sr_check_period、sr_check_user、sr_check_password 后，我在新日志中遇到的错误，现在所有以前的错误都消失了：

2016-03-31 17:45:00: pid 18363: DEBUG:  detect error: kind: 1
2016-03-31 17:45:00: pid 18363: DEBUG:  reading backend data packet kind
2016-03-31 17:45:00: pid 18363: DETAIL:  backend:0 of 2 kind = '1'
...
2016-03-31 17:45:00: pid 18363: DEBUG:  detect error: kind: S

Answer 1

故障转移脚本未执行的原因可能有多种。主要步骤是将 log_destination 属性启用到 syslog 并启用调试模式 (debug_level =1) 。

我目睹了故障转移脚本无法获取 %d、%H 的参数（特殊字符）的场景，因为脚本无法通过 ssh 连接到从站并触摸触发器文件。

如果您发布相同的日志文件，我可以提供更多详细信息。

根据新的日志： 我可以看到一个 ERROR: failed to authenticate 。 能否检查一下pgpool的以下参数是否配置正确

health_check_user
健康检查密码
恢复用户
恢复密码
wd_lifecheck_user
wd_lifecheck_password
sr_check_user
sr_check_password

希望你已经按照修改 postgres 用户密码的步骤操作了

alter user postgres password 'yourpassword'

并确保在所有情况下都提供相同的密码。

从日志来看，这似乎是一个身份验证问题。你能告诉我你正在使用的 pgpool 版本吗？

这些是我们用于设置 3 台机器（1 台主机、1 台从机和 1 台机器用于 pgpool）的配置 我已修改以适合您的 IP 地址

 listen_addresses = '*'
  port = 5433
  socket_dir = '/var/run/postgresql'
  pcp_port = 9898
  pcp_socket_dir = '/var/run/postgresql'

  backend_hostname0 = '192.168.0.2'
  backend_port0 = 5432
  backend_weight0 = 1
  backend_data_directory0 = '/var/lib/postgresql/9.4/main'
  backend_flag0 = 'ALLOW_TO_FAILOVER'

  backend_hostname1 = '192.168.0.3'
  backend_port1 = 5432
  backend_weight1 = 1
  backend_data_directory1 = '/var/lib/postgresql/9.4/main'
  backend_flag1 = 'ALLOW_TO_FAILOVER'

  enable_pool_hba = on
  pool_passwd = ''
  authentication_timeout = 60
  ssl = off
  num_init_children = 4
  max_pool = 2
  child_life_time = 300 
  child_max_connections = 0
  connection_life_time = 0
  client_idle_limit = 0
  log_destination = 'stderr,syslog'
  print_timestamp = on
  log_connections = on
  log_hostname = on
  log_statement = on
  log_per_node_statement = on
  log_standby_delay = 'none'
  syslog_facility = 'LOCAL0'
  syslog_ident = 'pgpool'
  debug_level = 1
  pid_file_name = '/var/run/postgresql/pgpool.pid'
  logdir = '/var/log/postgresql'
  connection_cache = on
  reset_query_list = 'ABORT; DISCARD ALL'

  replication_mode = off
  replicate_select = off
  insert_lock = on
  lobj_lock_table = ''
  replication_stop_on_mismatch = off
  failover_if_affected_tuples_mismatch = off

  load_balance_mode = off
  ignore_leading_white_space = on
  white_function_list = ''
  black_function_list = 'nextval,setval'

  master_slave_mode = on
  master_slave_sub_mode = 'stream'
  sr_check_period = 10
  sr_check_user = 'postgres'
  sr_check_password = 'postgres123'
  delay_threshold = 0
  follow_master_command = ''
  parallel_mode = off
  pgpool2_hostname = 'pgmaster'

  system_db_hostname  = 'localhost'
  system_db_port = 5432
  system_db_dbname = 'pgpool'
  system_db_schema = 'pgpool_catalog'
  system_db_user = 'pgpool'
  system_db_password = ''

  health_check_period = 5
  health_check_timeout = 20
  health_check_user = 'postgres'
  health_check_password = 'postgres123'
  health_check_max_retries = 2
  health_check_retry_delay = 1

  failover_command = '/usr/sbin/failover_modified.sh %d "%H" %P /var/lib/postgresql/9.4/main/pgsql.trigger.5432'
  failback_command = ''
  fail_over_on_backend_error = on
  search_primary_node_timeout = 10

  recovery_user = 'postgres'
  recovery_password = 'postgres123'
  recovery_1st_stage_command = ''
  recovery_2nd_stage_command = ''
  recovery_timeout = 90
  client_idle_limit_in_recovery = 0

  use_watchdog = off
  trusted_servers = ''
  ping_path = '/bin'
  wd_hostname = ''
  wd_port = 9000
  wd_authkey = ''
  delegate_IP = ''
  ifconfig_path = '/sbin'
  if_up_cmd = 'ifconfig eth0:0 inet $_IP_$ netmask 255.255.255.0'
  if_down_cmd = 'ifconfig eth0:0 down'
  arping_path = '/usr/sbin'  
  arping_cmd = 'arping -U $_IP_$ -w 1'

  clear_memqcache_on_escalation = on
  wd_escalation_command = ''

  wd_lifecheck_method = 'heartbeat'
  wd_interval = 10
  wd_heartbeat_port = 9694
  wd_heartbeat_keepalive = 2
  wd_heartbeat_deadtime = 30
  heartbeat_destination0 = '192.168.0.2'
  heartbeat_destination_port0 = 9694
  heartbeat_device0 = ''

  heartbeat_destination1 = '192.168.0.3'
  wd_life_point = 3
  wd_lifecheck_query = 'SELECT 1'
  wd_lifecheck_dbname = 'postgres'
  wd_lifecheck_user = 'postgres'
  wd_lifecheck_password = 'postgres123'

  relcache_expire = 0
  relcache_size = 256
  check_temp_table = on

  memory_cache_enabled = off
  memqcache_method = 'shmem'
  memqcache_memcached_host = 'localhost'
  memqcache_memcached_port = 11211
  memqcache_total_size = 67108864
  memqcache_max_num_cache = 1000000
  memqcache_expire = 0
  memqcache_auto_cache_invalidation = on
  memqcache_maxcache = 409600
  memqcache_cache_block_size = 1048576
  memqcache_oiddir = '/var/log/pgpool/oiddir'
  white_memqcache_table_list = ''
  black_memqcache_table_list = ''

另外，我希望您已经修改了 pool_hba.conf 以启用对主从的访问