Using io_method=worker in postgresql.conf file is causing psql client to waiting infinitely.

[VikasPatnala]

When io_method=worker in postgresql.conf file is causing psql client to waiting infinitely here is the demonstration

1. Initializing database
   • $ bin/initdb -D mydb
2. Starting the postgres server with debug option
   • $ bin/pg_ctl -D mydb -l logfile start -o "-d 5"
3. Starting the psql client
   • $ bin/psql -d postgres

This will cause psql to wait infinitely.

Here is the stack trace of the both the bin/pg_ctl and bin/psql commands after the above commands explain where these commands are waiting

1. bin/psql -d postgres stack trace -> client stack trace

poll() at 0x2791b358, PPA1 at 0x27922238, stack at 0x50082fe6a0 
PQsocketPoll() at 0x272039c8, PPA1 at 0x272047c1, stack at 0x50082fe860 
pqSocketCheck() at 0x27203678, PPA1 at 0x27204771, stack at 0x50082fe980 
pqWaitTimed() at 0x27203528, PPA1 at 0x27204749, stack at 0x50082feb80 
pqConnectDBComplete() at 0x271ebb30, PPA1 at 0x271ffc18, stack at 0x50082fec80 
PQconnectdbParams() at 0x271eb710, PPA1 at 0x271ffbbc, stack at 0x50082feda0 
main() at 0x27191750, PPA1 at 0x27196e10, stack at 0x50082feea0 
CELQINIT() at 0x27298010, PPA1 at 0x2729a024, stack at 0x50082ff200 

stopped in poll at 0x000000002791b5c6 ($t1) 
0x000000002791b5c6 (???) 0706 BCR 0,R6

2. bin/pg_ctl -D mydb -l logfile start -> server stack trace

FDBX0254: Stack trace follows
poll() at 0x28b90358, PPA1 at 0x28b97238, stack at 0x50082fe1c0
WaitEventSetWaitBlock() at 0x27d1cc70, PPA1 at 0x27d1dd8a, stack at 0x50082fe380
WaitEventSetWait() at 0x27d1c810, PPA1 at 0x27d1dcfa, stack at 0x50082fe4a0
ServerLoop() at 0x27b81910, PPA1 at 0x27b8a99f, stack at 0x50082fe5c0
PostmasterMain() at 0x27b7ccd0, PPA1 at 0x27b8a5d3, stack at 0x50082fed60
main() at 0x27901780, PPA1 at 0x2790328e, stack at 0x50082ff000
CELQINIT() at 0x2850d010, PPA1 at 0x2850f024, stack at 0x50082ff200

stopped in poll at 0x0000000028b905c6 ($t1)
0x0000000028b905c6 (???)     0706         BCR     0,R6

[VikasPatnala]

I had a debug statement in PQsocketPoll for client side and WaitEventSetWaitBlock for server side to check the socket they are listening to is same or not?

Here is the output from client

/home/sravani/postgresbin>bin/psql -d postgres
28Sep Client-side debug, sock=3, forRead=0, forWrite=1
28Sep Client-side debug, sock=3, forRead=1, forWrite=0

output from server

28Sep-Debug, cur_timeout=200, sock=3
28Sep-Debug, cur_timeout=200, sock=3
...repeating the above line

from the above logs here are few observations

• Client sends a connection request to the server and waits for the response
• Since it's repeating the try for poll, it's not receiving the client's connection request or is unable to process it.

[VikasPatnala]

In the logfile which is generated with debug info when we run the server contains the line

2025-09-28 13:34:44.682 EDT [50333405] DEBUG:  forked new client backend, pid=3535 socket=10
2025-09-28 13:34:44.683 EDT [3535] LOG:  connection received: host=[local]

which confirms that the server's postmaster process accepted the connection and a new backend process (the one that handles the actual client session) has been spawned or assigned to handle this connection.

logfile_zos.txt

but I don't see the below lines which should come in case of success

(this log is taken from the mac system)
2025-09-26 20:34:54.278 IST [35890] LOG:  connection authenticated: user="vikaspatnala" method=trust (/Users/vikaspatnala/postgresport/postgresbin/mydb/pg_hba.conf:117)
2025-09-26 20:34:54.278 IST [35890] LOG:  connection authorized: user=vikaspatnala database=postgres application_name=psql