GaussDB(DWS)集群后台UDF进程异常

背景：使用UDF出现报错，后台查看集群UDF进程发现进程异常。本帖通过简单示例介绍UDF进程异常的排查和处理方式，示例中用到的数据、路径配置、主机名等信息均为测试环境信息。

首先，通过以下命令查看集群UDF进程状态：

cm_ctl query -CvF

结果如下：

[  Fenced UDF State   ]

node         state
--------------------
1  ASG003    Down
2  host17967 Down
3  host17995 Down

发现UDF进程处于异常状态，查看cm_agent日志（在$GAUSSLOG/cm/cm_agent目录下）发现如下内容：

StartAndStop LOG: process (secbox) is not running, path is xxxx, have_found is 0

该日志说明secbox.conf的配置存在问题，进入$GAUSSHOME/secbox路径下，查看secbox.conf配置：

# read/write src_path [dst_path]
[mount_path]            read    /dev
[mount_path]            read    /sys
[mount_path]            read    /bin
[mount_path]            read    /sbin
[mount_path]            read    /usr/bin
[mount_path]            read    /lib
[mount_path]            read    /lib64
[mount_path]            read    /usr/lib
[mount_path]            read    /usr/lib64
[mount_path]            read    /usr/local
[mount_path]            read    /usr/share
[mount_path]            read    /etc
[mount_path]            read    /var
[mount_path]            read    /var/log
[mount_path]            read    /var/lkp

发现该配置中/var/lkp路径不存在，导致UDF进程拉起失败，注释或删除此条配置，几秒后观察本节点UDF进程恢复为Normal状态：

[  Fenced UDF State   ]

node         state
--------------------
1  ASG003    Down
2  host17967 Normal
3  host17995 Down

说明配置已生效，按照同样的方法对其他节点修改secbox.conf文件，问题解决。

[  Fenced UDF State   ]

node         state
--------------------
1  ASG003    Normal
2  host17967 Normal
3  host17995 Normal