因 pg_hba.confg 中参数配置问题导致启动或重启集群失败的问题

一、问题现象

若openGauss资源池化集群启动或重启失败，有如下报错信息：

cm_ctl: start cluster failed in (601)s!

HINT: Maybe the cluster is continually being started in the background.

You can wait for a while and check whether the cluster starts, or increase the value of parameter "-t", e.g -t 600.

报告启动集群失败，10min 超时。

使用cm_ctl query -Cvipd查询集群状态后显示CMServer State中所有节点为正常，Datanode State中有的节点的状态为Down。

二、定位方法

登录故障节点机器，进入$PGDATA目录下，查看DBstart.log日志，发现如下几种报错信息：

2024-10-11 14:27:29.056 6708c550.1 [unknown] 281459392315408 [unknown] 0 dn_6001_6002 F0000  0 [BACKEND] LOG:  end-of-line before authentication method

2024-10-11 14:27:29.056 6708c550.1 [unknown] 281459392315408 [unknown] 0 dn_6001_6002 F0000  0 [BACKEND] CONTEXT:  line 94 of configuration file "/.../.../dn1/pg_hba.conf"

2024-10-11 14:27:29.056 6708c550.1 [unknown] 281459392315408 [unknown] 0 dn_6001_6002 42809  0 [BACKEND] FATAL:  could not load pg_hba.conf

2024-10-11 14:46:31.121 6708c9c6.1 [unknown] 281464592334864 [unknown] 0 dn_6001_6002 F0000  0 [BACKEND] LOG:  invalid authentication method "sm9"

2024-10-11 14:46:31.121 6708c9c6.1 [unknown] 281464592334864 [unknown] 0 dn_6001_6002 F0000  0 [BACKEND] CONTEXT:  line 94 of configuration file "/.../.../dn1/pg_hba.conf"

2024-10-11 14:46:31.121 6708c9c6.1 [unknown] 281464592334864 [unknown] 0 dn_6001_6002 42809  0 [BACKEND] FATAL:  could not load pg_hba.conf

通过上述这两种报错信息可以看出是pg_hba.conf文件内第 94 行 method 没有设置或设置的认证方式暂不支持。

进入$GAUSSLOG/pg_log/dn_xxx目录下，寻找该节点最近时间点的postgresql-xxx.log日志，发现如下报错信息：

2024-10-11 16:14:42.190 6708de72.5065 postgres 281464533200800 dn_6002 0 dn_6001_6002 28000  0 [BACKEND] FATAL:  no pg_hba.conf entry for host "::1", user "xxx", database "postgres", SSL on

2024-10-11 16:14:42.216 6708de72.5065 postgres 281464533200800 dn_6002 0 dn_6001_6002 28000  0 [BACKEND] FATAL:  no pg_hba.conf entry for host "::1", user "xxx", database "postgres", SSL off

2024-10-11 16:14:42.233 6708de34.6257 [unknown] 281464591396768 dn_6002 0 dn_6001_6002 YY004  0 [BACKEND] ERROR:  Could not connect to the postgres, we have tried 10 times, the connection info: FATAL:  no pg_hba.conf entry for host "::1", user "xxx", database "postgres", SSL on

        FATAL:  no pg_hba.conf entry for host "::1", user "xxx", database "postgres", SSL off

进入$GAUSSLOG/cm/cm_agent目录下，寻找该节点最近时间点的cm_agent日志，发现如下报错信息：

2024-10-11 16:49:55.955 tid=2425227  ERROR: get connect failed for dn(/.../.../dn1/postmaster.pid) phony dead check, errmsg is FATAL:  no pg_hba.conf entry for host "::1", user "xxx", database "postgres", SSL off

FATAL:  no pg_hba.conf entry for host "::1", user "xxx", database "postgres", SSL off

2024-10-11 16:49:55.955 tid=2425227  LOG: has found 51 times for instance(dn_6002) phony dead check.

2024-10-11 16:49:56.431 tid=2425223  ERROR: failed to connect to datanode:/.../.../dn1

2024-10-11 16:49:56.431 tid=2425223  ERROR: connection return errmsg : FATAL:  no pg_hba.conf entry for host "::1", user "xxx", database "postgres", SSL off

FATAL:  no pg_hba.conf entry for host "::1", user "xxx", database "postgres", SSL off

通过第二、三中报错信息可以看出是pg_hba.conf文件内 ip 设置异常导致的启动或重启集群失败。

三、问题根因

系统配置文件pg_hba.conf内各种参数出现异常导致的启动或重启集群失败。具体参数说明参考配置文件参考。

四、解决方案

针对该问题的解决方案为，查看DBstart.log和postgresql-xxx.log日志找到具体报错的参数，修改报错的参数为正常值或推荐值，然后重启集群即可。