Make the pg_rewind regression tests more robust on slow systems.

There were a couple of hard-coded sleeps in the tests: to wait for standby
to catch up with master, and to wait for promotion with "pg_ctl promote"
to complete. Instead of a fixed, hard-coded sleep, poll the server with a
query once a second. This isn't ideal either, and I wish we had a better
solution for real-world applications too, but this should fix the
immediate problem.

Patch by Michael Paquier, with some editing by me.

diff --git a/src/bin/pg_rewind/RewindTest.pm b/src/bin/pg_rewind/RewindTest.pm
index 84e0512143a34a007f84ea72074663e888baa3be..a5f7d08bf72afd6d8bfdc3398f703eaa989b2913 100644 (file)
--- a/src/bin/pg_rewind/RewindTest.pm
+++ b/src/bin/pg_rewind/RewindTest.pm
@@ -125,6 +125,37 @@ sub check_query
    }
 }
 
+# Run a query once a second, until it returns 't' (i.e. SQL boolean true).
+sub poll_query_until
+{
+   my ($query, $connstr) = @_;
+
+   my $max_attempts = 30;
+   my $attempts = 0;
+   my ($stdout, $stderr);
+
+   while ($attempts < $max_attempts)
+   {
+       my $cmd = ['psql', '-At', '-c', "$query", '-d', "$connstr" ];
+       my $result = run $cmd, '>', \$stdout, '2>', \$stderr;
+
+       chomp($stdout);
+       if ($stdout eq "t")
+       {
+           return 1;
+       }
+
+       # Wait a second before retrying.
+       sleep 1;
+       $attempts++;
+   }
+
+   # The query result didn't change in 30 seconds. Give up. Print the stderr
+   # from the last attempt, hopefully that's useful for debugging.
+   diag $stderr;
+   return 0;
+}
+
 sub append_to_file
 {
    my($filename, $str) = @_;
@@ -185,7 +216,7 @@ sub create_standby
    # Base backup is taken with xlog files included
    system_or_bail("pg_basebackup -D $test_standby_datadir -p $port_master -x >>$log_path 2>&1");
    append_to_file("$test_standby_datadir/recovery.conf", qq(
-primary_conninfo='$connstr_master'
+primary_conninfo='$connstr_master application_name=rewind_standby'
 standby_mode=on
 recovery_target_timeline='latest'
 ));
@@ -193,8 +224,11 @@ recovery_target_timeline='latest'
    # Start standby
    system_or_bail("pg_ctl -w -D $test_standby_datadir -o \"-k $tempdir_short --listen-addresses='' -p $port_standby\" start >>$log_path 2>&1");
 
-   # sleep a bit to make sure the standby has caught up.
-   sleep 1;
+   # Wait until the standby has caught up with the primary, by polling
+   # pg_stat_replication.
+   my $caughtup_query = "SELECT pg_current_xlog_location() = replay_location FROM pg_stat_replication WHERE application_name = 'rewind_standby';";
+   poll_query_until($caughtup_query, $connstr_master)
+       or die "Timed out while waiting for standby to catch up";
 }
 
 sub promote_standby
@@ -203,9 +237,11 @@ sub promote_standby
    # up standby
 
    # Now promote slave and insert some new data on master, this will put
-   # the master out-of-sync with the standby.
+   # the master out-of-sync with the standby. Wait until the standby is
+   # out of recovery mode, and is ready to accept read-write connections.
    system_or_bail("pg_ctl -w -D $test_standby_datadir promote >>$log_path 2>&1");
-   sleep 2;
+   poll_query_until("SELECT NOT pg_is_in_recovery()", $connstr_standby)
+       or die "Timed out while waiting for promotion of standby";
 }
 
 sub run_pg_rewind