Enhance native replication mode or snapshot isolation mode online

recovery case.

Use streaming replication in the first stage and replays WAL generated
during base copy as much as possible. In the second stage, now it just
promotes the standby server. If there are little or zero WAL to
replay, this will significantly reduce the duration of the second
stage. This is important because while the system in the second stage,
frontend cannot connect to Pgpool-II.

diff --git a/src/test/pgpool_setup.in b/src/test/pgpool_setup.in
index 5d04ed5387caa817a0ba2ea855513aa0ed37e0b6..dfc767c619a128b0712b21360def63e031743e44 100644 (file)
--- a/src/test/pgpool_setup.in
+++ b/src/test/pgpool_setup.in
@@ -414,8 +414,11 @@ cat >> $1/$SCRIPT <<'EOF'
 #! /bin/sh
 psql=__PGBIN__/psql
 pg_rewind=__PGBIN__/pg_rewind
+PG_CTL=__PGBIN__/pg_ctl
 DATADIR_BASE=__DATADIR_BASE__
 PGSUPERUSER=__PGSUPERUSER__
+MAX_DURATION=60
+pgversion=__PGVERSION__
 
 main_db_cluster=$1
 recovery_node_host_name=$2
@@ -532,80 +535,56 @@ EOF
        -e "/__PGBIN__/s@__PGBIN__@$PGBIN@" \
        -e "/__DATADIR_BASE__/s@__DATADIR_BASE__@$BASEDIR@" \
        -e "/__PGSUPERUSER__/s/__PGSUPERUSER__/$WHOAMI/" \
+       -e "/__PGVERSION__/s/__PGVERSION__/$PGVERSION/" \
 $1/$SCRIPT
 
 chmod 755 $1/$SCRIPT
 }
 
 #-------------------------------------------
-# create basebackup.sh for native replication mode
-# argument: PostgreSQL database cluster directory
+# create basebackup.sh for native replication mode and snapshot isolation mode.
+# argument1: full path to database cluster directory of main node
+# argument2: node number of target node
 #-------------------------------------------
 function create_basebackup_replication {
-SCRIPT=basebackup.sh
-cat >> $1/$SCRIPT <<'EOF'
-#! /bin/sh
-psql=__PGBIN__/psql
-DATADIR_BASE=__DATADIR_BASE__
-PGSUPERUSER=__PGSUPERUSER__
-EOF
-
-echo "export PGHOST=$PGSOCKET_DIR" >> $1/$SCRIPT
-
-cat >> $1/$SCRIPT <<'EOF'
-main_db_cluster=$1
-recovery_node_host_name=$2
-DEST_CLUSTER=$3
-PORT=$4
-
-log=$DATADIR_BASE/log/recovery.log
+    CLUSTERDIR=$1
+    n=$2
+    SCRIPT=basebackup.sh
 
-$psql -p $PORT -c "SELECT pg_start_backup('Native Replication', true)" postgres
-
-echo "source: $main_db_cluster dest: $DEST_CLUSTER" > $log
-EOF
-
-if [ $PGVERSION -ge 120 ];then
+    create_basebackup_stream $CLUSTERDIR $n
     cat >> $1/$SCRIPT <<'EOF'
-cat > $main_db_cluster/myrecovery.conf <<REOF
-restore_command = 'cp __ARCHDIR__/%f %p'
-REOF
-EOF
-
-else
-
-    cat >> $1/$SCRIPT <<'EOF'
-cat > $main_db_cluster/recovery.conf <<REOF
-restore_command = 'cp __ARCHDIR__/%f %p'
-REOF
-EOF
-fi
-
-cat >> $1/$SCRIPT <<'EOF'
-rsync -C -a -c --delete --exclude postgresql.conf --exclude postmaster.pid \
---exclude postmaster.opts --exclude pg_log \
---exclude recovery.done \
---exclude pg_xlog pg_wal \
-$main_db_cluster/ $DEST_CLUSTER/
+# start target server as a streaming replication standby server
+$PG_CTL -w -D $DEST_CLUSTER start
 
-rm -fr $DEST_CLUSTER/pg_xlog 
-mkdir $DEST_CLUSTER/pg_xlog
-chmod 700 $DEST_CLUSTER/pg_xlog
-
-rm -fr $DEST_CLUSTER/pg_wal
-mkdir $DEST_CLUSTER/pg_wal
-chmod 700 $DEST_CLUSTER/pg_wal
-EOF
+# wait till the standby catches up primary server or
+# $MAX_DURATION seconds passes
+sec=0
+while [ $sec -lt $MAX_DURATION ]
+do
+  sec=`expr $sec + 1`
+
+  if [ $pgversion -ge 100 ];then
+    p_lsn=`psql -p $4 -c "SELECT pg_current_wal_lsn()" postgres|sed -n 3p|sed 's/ //'`
+    s_lsn=`psql -p $6 -c "SELECT pg_last_wal_replay_lsn()" postgres|sed -n 3p|sed 's/ //'`
+  else
+    p_lsn=`psql -p $4 -c "SELECT pg_current_xlog_lsn()" postgres|sed -n 3p|sed 's/ //'`
+    s_lsn=`psql -p $6 -c "SELECT pg_last_xlog_replay_lsn()" postgres|sed -n 3p|sed 's/ //'`
+  fi
 
-cat >> $1/$SCRIPT <<'EOF'
-$psql -p $PORT -c "SELECT pg_stop_backup()" postgres
-EOF
+  if [ $pgversion -ge 100 ];then
+    sql=`echo "SELECT pg_wal_lsn_diff('$p_lsn','$s_lsn');"`
+  else
+    sql=`echo "SELECT pg_xlog_location_diff('$p_lsn','$s_lsn');"`
+  fi
 
-if [ $PGVERSION -ge 120 ];then
-    cat >> $1/$SCRIPT <<'EOF'
-touch $DEST_CLUSTER/recovery.signal
+  diff=`psql -p $4 -c "$sql" postgres|sed -n 3p|sed 's/ //g'`
+#  echo "p_lsn: $p_lsn s_lsn: $s_lsn diff: $diff"
+  if [ $diff -le 0 ];then
+    sec=$MAX_DURATION
+  fi
+  sleep 1
+done
 EOF
-fi
 
 #-------------------------------------------
 # replace some variables in the script
@@ -615,7 +594,8 @@ fi
        -e "/__DATADIR_BASE__/s@__DATADIR_BASE__@$BASEDIR@" \
        -e "/__PGSUPERUSER__/s/__PGSUPERUSER__/$WHOAMI/" \
        -e "/__ARCHDIR__/s@__ARCHDIR__@$BASEDIR/archivedir/\`basename \$1\`@" \
-       $1/$SCRIPT
+       -e "/__PGVERSION__/s/__PGVERSION__/$PGVERSION/" \
+$1/$SCRIPT
 
 chmod 755 $1/$SCRIPT
 }
@@ -631,6 +611,7 @@ cat >> $1/$SCRIPT <<'EOF'
 psql=__PGBIN__/psql
 DATADIR_BASE=__DATADIR_BASE__
 PGSUPERUSER=__PGSUPERUSER__
+PGCTL=__PGBIN__/pg_ctl
 
 main_db_cluster=$1
 recovery_node_host_name=$2
@@ -653,6 +634,8 @@ do
 done
 
 $psql -p $PORT -c "SELECT pgpool_switch_xlog('__ARCHDIR__')" template1
+
+$PGCTL -D $DEST_CLUSTER promote
 EOF
 #-------------------------------------------
 # replace some variables in the script
@@ -972,8 +955,8 @@ do
        if [ $MODE = 's' ];then
                create_basebackup_stream $CLUSTERDIR $n >&5 2>&1
        elif [ $MODE = 'r' -o $MODE = 'i' ];then
-               create_basebackup_replication $CLUSTERDIR >&5 2>&1
-               create_pgpool_recovery_pitr $CLUSTERDIR >&5 2>&1
+               create_basebackup_replication $CLUSTERDIR $n >&5 2>&1
+               create_pgpool_recovery_pitr $CLUSTERDIR $n >&5 2>&1
        fi
 
        # create recovery.conf or recovery.done if streaming replication