Hi All, Here is a strange behaviour with master branch with head at commit d3c4c471553265e7517be24bae64b81967f6df40 Author: Peter Eisentraut <pete...@gmx.net> Date: Mon Feb 10 21:47:19 2014 -0500
The OS is [ashutosh@ubuntu repro]uname -a Linux ubuntu 3.2.0-59-generic #90-Ubuntu SMP Tue Jan 7 22:43:51 UTC 2014 x86_64 x86_64 x86_64 GNU/Linux This is a VM hosted on Mac-OS X 10.7.5 Here's the SQL script [ashutosh@ubuntu repro]cat big_select_killed.sql drop table big_tab; create table big_tab (val int, val2 int, str varchar); insert into big_tab select x, x, lpad('string', 100, x::text) from generate_series(1, 10000000) x; select * from big_tab; The last select causes the "Killed" message. [ashutosh@ubuntu repro]psql -d postgres -f big_select_killed.sql DROP TABLE CREATE TABLE INSERT 0 10000000 Killed There is a message in server log FATAL: connection to client lost STATEMENT: select * from big_tab; Any SELECT selecting all the rows is getting psql killed but not SELECT count(*) [ashutosh@ubuntu repro]psql -d postgres psql (9.4devel) Type "help" for help. postgres=# select count(*) from big_tab; count ---------- 10000000 (1 row) postgres=# select * from big_tab; Killed [ashutosh@ubuntu repro]psql -d postgres psql (9.4devel) Type "help" for help. Below is the buffer cache size and the relation size (if anyone cares) postgres=# show shared_buffers; shared_buffers ---------------- 128MB (1 row) postgres=# select pg_relation_size('big_tab'::regclass); pg_relation_size ------------------ 1412415488 (1 row) postgres=# select pg_relation_size('big_tab'::regclass)/1024/1024; -- IN MBs to be simple ?column? ---------- 1346 (1 row) There are no changes in default configuration. Using unix sockets [ashutosh@ubuntu repro]ls /tmp/.s.PGSQL.5432* /tmp/.s.PGSQL.5432 /tmp/.s.PGSQL.5432.lock Looks like a bug in psql to me. Does anybody see that behaviour? -- Best Wishes, Ashutosh Bapat EnterpriseDB Corporation The Postgres Database Company