Postgres Socket file getting deleted

i was encountering a situation where my Postgres Unix socket File gets deleted . Ideally when our application is running the Postgres deamon should not go down at all . However the postmaster does go down and the unix socket file created at /tmp/ also gets deleted . This has serious implications for our products .

I wanted to know when can the socket file get deleted if one does not do it explicitly . Does it mean the Postgres connection has g开发者_JAVA百科one bad and it re-started itself ? Is that is the case when can the connection go bad ?

It's unclear if you mean whether the socket file gets removed because postgresql goes down (which is normal) or whether postgresql goes down because the socket file is removed. Check the postgres logs.

It could be there's a procedure that cleans up old files in /tmp/ . e.g. CentOS/RHEL does this by the means of tmpwatch , the /etc/cron.daily/tmpwatch script specifically. And although tmpwatch shouldn't remove the pgsql file, who knows if there's a bug lurking..