Test: attempt to fix 001.load_balance test timeout.
authorTatsuo Ishii <ishii@sraoss.co.jp>
Wed, 13 Sep 2023 08:51:21 +0000 (17:51 +0900)
committerTatsuo Ishii <ishii@sraoss.co.jp>
Wed, 13 Sep 2023 09:17:03 +0000 (18:17 +0900)
commitfb0800e3f0211ba833ca05cf506387e4d2a279b8
treea6075f0bafe910e1f52e516ff47bd91713134ed6
parentd8d288776a72be49badbbea111edba0223d597c0
Test: attempt to fix 001.load_balance test timeout.

Examining the log reveals that the error was caused by bind error:
address already used.  This is strange because SO_REUSEADDR flags is
set to the listening socket.  I guess the cause is, the kernel is not
fast enough to clean up the socket used by previous pgpool process.
To fix this, existing CHECK_TIME_WAIT environment variable parameter
is modified so that it checks the socket by using netstat -tulpn,
rather than netstat -a, which is too slow.
src/test/pgpool_setup.in
src/test/regression/tests/001.load_balance/test.sh