Hello,
I tried the following scenario on PostgreSQL 9.4.2 on Linux:
1) Connect to a remote postgres database and run a long running query
e.g. psql -h vm5 -p 57001 warehouse -c 'select x.a from
generate_series(1, 100000000) as x(a);'
2) Suspend the client process by pressing CTRL-Z
3) In another session, get the backend PID from pg_stat_activity and run
pg_terminate_backend(PID)
or
pg_cancel_backend(PID)
The query hangs indeterminately and cannot be terminated.
Using strace, I see that the backend is stuck trying to send to the client
in the function internal_flush(), in send(). The backend does receive the
SIGINT or SIGTERM signal, but it is just restarting the system call.
sendto(9, "D\0\0\0\22\0\1\0\0\0\01017182365D\0\0\0\22\0\1\0\0\0\01017"...,
5392, 0, NULL, 0) = ? ERESTARTSYS (To be restarted)
--- SIGINT (Interrupt) @ 0 (0) ---
rt_sigreturn(0x2b85fc4234fc) = 44
sendto(9, "D\0\0\0\22\0\1\0\0\0\01017182365D\0\0\0\22\0\1\0\0\0\01017"...,
5392, 0, NULL, 0) = ? ERESTARTSYS (To be restarted)
--- SIGTERM (Terminated) @ 0 (0) ---
rt_sigreturn(0x2b85fc4234fc) = 44
sendto(9, "D\0\0\0\22\0\1\0\0\0\01017182365D\0\0\0\22\0\1\0\0\0\01017"...,
5392, 0, NULL, 0) = ? ERESTARTSYS (To be restarted)
We have several customers that see these "hanging" queries and have no way
of canceling them, and I suspect that this is the cause.
Has anyone else encountered this? Is there a way to cancel these queries,
or fix the problem in the code?
Thanks,
George
--
Sent via pgsql-general mailing list (pgsql-general@xxxxxxxxxxxxxx)
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-general
