The Control-M Server Postgres database crashes with error message “sorry, too many clients already” Several messages can be seen in databases logs EDTFATAL: sorry, too many clients already EDTFATAL: sorry, too many clients already EDTFATAL: sorry, too many clients already EDTFATAL: sorry, too many clients already EDTFATAL: sorry, too many clients already Control-M Server may continue working for some time but eventually database will become unresponsive and crashes, it's necessary to recycle the host machine to have the database working again |
CTM-3653 and CTM-9416 were created for this issue and the problem is resolved when upgrading PG to 15.3 that supports memory type “sysv”. if Dedicated PG is used then upgrade Control-M/Server to 9.0.21.200 and then upgrade PG to 15.3. Add the line: shared_memory_type=sysv to postgresql.conf file • Requires DB restart • PG 10.4, unlike previous versions comes bundled with shared_memory_type “mmap” • AIX cannot really support “mmap” for very frequest logins/logouts. It requires the value of “sysv”.which is supported from PG 15.3 |