galera SST scripts: unification of the previous SST completion check

scripts/wsrep_sst_backup.sh

@@ -22,14 +22,11 @@ set -ue
 # This is a reference script for rsync-based state snapshot transfer
 
 . $(dirname "$0")/wsrep_sst_common
-wsrep_check_datadir
-
-DATA="$WSREP_SST_OPT_DATA"
-
-create_data
 
 MAGIC_FILE="$DATA/backup_sst_complete"
 
+wait_previous_sst
+
 [ -f "$MAGIC_FILE" ] && rm -f "$MAGIC_FILE"
 
 if [ "$WSREP_SST_OPT_ROLE" = 'donor' ]

scripts/wsrep_sst_common.sh

@@ -1769,7 +1769,7 @@ simple_cleanup()
     if [ $estatus -ne 0 ]; then
         wsrep_log_error "Cleanup after exit with status: $estatus"
     fi
-    if [ -n "${SST_PID:-}" ]; then
+    if [ -n "$SST_PID" ]; then
         [ "$(pwd)" != "$OLD_PWD" ] && cd "$OLD_PWD"
         [ -f "$SST_PID" ] && rm -f "$SST_PID" || :
     fi
@@ -1869,4 +1869,29 @@ create_dirs()
     fi
 }
 
+wait_previous_sst()
+{
+    # give some time for previous SST to complete:
+    check_round=0
+    while check_pid "$SST_PID" 1; do
+        wsrep_log_info "Previous SST is not completed, waiting for it to exit"
+        check_round=$(( check_round+1 ))
+        if [ $check_round -eq 30 ]; then
+            wsrep_log_error "previous SST script still running..."
+            exit 114 # EALREADY
+        fi
+        sleep 1
+    done
+
+    trap simple_cleanup EXIT
+    echo $$ > "$SST_PID"
+}
+
+DATA="$WSREP_SST_OPT_DATA"
+
+wsrep_check_datadir
+create_data
+
+SST_PID="$DATA/wsrep_sst.pid"
+
 wsrep_log_info "$WSREP_METHOD $WSREP_TRANSFER_TYPE started on $WSREP_SST_OPT_ROLE"

scripts/wsrep_sst_mariabackup.sh

@@ -24,7 +24,21 @@ set -ue
 # Make sure to read that before proceeding!
 
 . $(dirname "$0")/wsrep_sst_common
-wsrep_check_datadir
+
+BACKUP_BIN=$(commandex 'mariadb-backup')
+if [ -z "$BACKUP_BIN" ]; then
+    wsrep_log_error 'mariadb-backup binary not found in path'
+    exit 42
+fi
+
+BACKUP_PID=""
+
+INFO_FILE='xtrabackup_galera_info'
+DONOR_INFO_FILE='donor_galera_info'
+IST_FILE='xtrabackup_ist'
+
+MAGIC_FILE="$DATA/$INFO_FILE"
+DONOR_MAGIC_FILE="$DATA/$DONOR_INFO_FILE"
 
 ealgo=""
 eformat=""
@@ -33,7 +47,6 @@ ekeyfile=""
 encrypt=0
 ssyslog=""
 ssystag=""
-BACKUP_PID=""
 tcert=""
 tcap=""
 tpem=""
@@ -62,10 +75,10 @@ tcmd=""
 payload=0
 pvformat="-F '%N => Rate:%r Avg:%a Elapsed:%t %e Bytes: %b %p'"
 pvopts="-f -i 10 -N $WSREP_SST_OPT_ROLE"
-STATDIR=""
 uextra=0
 disver=""
 
+STATDIR=""
 tmpopts=""
 itmpdir=""
 xtmpdir=""
@@ -91,23 +104,6 @@ readonly TOTAL_TAG='total'
 # For backup locks it is 1 sent by joiner
 sst_ver=1
 
-declare -a RC
-
-BACKUP_BIN=$(commandex 'mariadb-backup')
-if [ -z "$BACKUP_BIN" ]; then
-    wsrep_log_error 'mariadb-backup binary not found in path'
-    exit 42
-fi
-
-DATA="$WSREP_SST_OPT_DATA"
-
-INFO_FILE='xtrabackup_galera_info'
-DONOR_INFO_FILE='donor_galera_info'
-IST_FILE='xtrabackup_ist'
-
-MAGIC_FILE="$DATA/$INFO_FILE"
-DONOR_MAGIC_FILE="$DATA/$DONOR_INFO_FILE"
-
 INNOAPPLYLOG="$DATA/mariabackup.prepare.log"
 INNOMOVELOG="$DATA/mariabackup.move.log"
 INNOBACKUPLOG="$DATA/mariabackup.backup.log"
@@ -713,7 +709,7 @@ cleanup_at_exit()
     fi
 
     # Final cleanup
-    pgid=$(ps -o 'pgid=' $$ 2>/dev/null | grep -o -E '[0-9]+' || :)
+    local pgid=$(ps -o 'pgid=' $$ 2>/dev/null | grep -o -E '[0-9]+' || :)
 
     # This means no setsid done in mysqld.
     # We don't want to kill mysqld here otherwise.
@@ -727,7 +723,7 @@ cleanup_at_exit()
         fi
     fi
 
-    if [ -n "${SST_PID:-}" ]; then
+    if [ -n "$SST_PID" ]; then
         [ -f "$SST_PID" ] && rm -f "$SST_PID" || :
     fi
 
@@ -933,7 +929,17 @@ if "$BACKUP_BIN" --help 2>/dev/null | grep -qw -F -- '--version-check'; then
     disver=' --no-version-check'
 fi
 
-create_data
+get_stream
+get_transfer
+
+findopt='-L'
+[ "$OS" = 'FreeBSD' ] && findopt="$findopt -E"
+
+wait_previous_sst
+
+[ -f "$MAGIC_FILE" ] && rm -f "$MAGIC_FILE"
+[ -f "$DONOR_MAGIC_FILE" ] && rm -f "$DONOR_MAGIC_FILE"
+[ -f "$DATA/$IST_FILE" ] && rm -f "$DATA/$IST_FILE"
 
 if [ $ssyslog -eq 1 ]; then
     if [ -n "$(commandex logger)" ]; then
@@ -1053,30 +1059,6 @@ send_magic()
     fi
 }
 
-get_stream
-get_transfer
-
-findopt='-L'
-[ "$OS" = 'FreeBSD' ] && findopt="$findopt -E"
-
-SST_PID="$DATA/wsrep_sst.pid"
-
-# give some time for previous SST to complete:
-check_round=0
-while check_pid "$SST_PID"; do
-    wsrep_log_info "previous SST is not completed, waiting for it to exit"
-    check_round=$(( check_round+1 ))
-    if [ $check_round -eq 30 ]; then
-       wsrep_log_error "previous SST script still running."
-       exit 114 # EALREADY
-    fi
-    sleep 1
-done
-
-[ -f "$MAGIC_FILE" ] && rm -f "$MAGIC_FILE"
-[ -f "$DONOR_MAGIC_FILE" ] && rm -f "$DONOR_MAGIC_FILE"
-[ -f "$DATA/$IST_FILE" ] && rm -f "$DATA/$IST_FILE"
-
 if [ "$WSREP_SST_OPT_ROLE" = 'donor' ]; then
 
     trap cleanup_at_exit EXIT
@@ -1249,9 +1231,6 @@ else # joiner
         impts="--parallel=$backup_threads${impts:+ }$impts"
     fi
 
-    trap simple_cleanup EXIT
-    echo $$ > "$SST_PID"
-
     stagemsg='Joiner-Recv'
 
     MODULE="${WSREP_SST_OPT_MODULE:-xtrabackup_sst}"

scripts/wsrep_sst_mysqldump.sh

@@ -23,6 +23,8 @@ set -ue
 
 . $(dirname "$0")/wsrep_sst_common
 
+wait_previous_sst
+
 EINVAL=22
 
 if test -z "$WSREP_SST_OPT_HOST";  then wsrep_log_error "HOST cannot be nil";  exit $EINVAL; fi

scripts/wsrep_sst_rsync.sh

@@ -22,13 +22,22 @@ set -ue
 # This is a reference script for rsync-based state snapshot transfer
 
 . $(dirname "$0")/wsrep_sst_common
-wsrep_check_datadir
 
 wsrep_check_programs rsync
 
 RSYNC_REAL_PID=0   # rsync process id
 STUNNEL_REAL_PID=0 # stunnel process id
 
+MODULE="${WSREP_SST_OPT_MODULE:-rsync_sst}"
+
+RSYNC_PID="$DATA/$MODULE.pid"
+RSYNC_CONF="$DATA/$MODULE.conf"
+
+STUNNEL_CONF="$DATA/stunnel.conf"
+STUNNEL_PID="$DATA/stunnel.pid"
+
+MAGIC_FILE="$DATA/rsync_sst_complete"
+
 cleanup_joiner()
 {
     # Since this is invoked just after exit NNN
@@ -150,13 +159,6 @@ check_pid_and_port()
     check_pid "$pid_file" && [ $CHECK_PID -eq $pid ]
 }
 
-DATA="$WSREP_SST_OPT_DATA"
-
-STUNNEL_CONF="$DATA/stunnel.conf"
-STUNNEL_PID="$DATA/stunnel.pid"
-
-MAGIC_FILE="$DATA/rsync_sst_complete"
-
 get_binlog
 
 if [ -n "$WSREP_SST_OPT_BINLOG" ]; then
@@ -164,8 +166,6 @@ if [ -n "$WSREP_SST_OPT_BINLOG" ]; then
     binlog_base=$(basename "$WSREP_SST_OPT_BINLOG")
 fi
 
-create_data
-
 BINLOG_TAR_FILE="$DATA_DIR/wsrep_sst_binlog.tar"
 
 ar_log_dir="$DATA_DIR"
@@ -270,22 +270,7 @@ fi
 readonly SECRET_TAG='secret'
 readonly BYPASS_TAG='bypass'
 
-SST_PID="$DATA/wsrep_sst.pid"
+wait_previous_sst
-
-# give some time for previous SST to complete:
-check_round=0
-while check_pid "$SST_PID"; do
-    wsrep_log_info "Previous SST is not completed, waiting for it to exit"
-    check_round=$(( check_round+1 ))
-    if [ $check_round -eq 20 ]; then
-        wsrep_log_error "previous SST script still running."
-        exit 114 # EALREADY
-    fi
-    sleep 1
-done
-
-trap simple_cleanup EXIT
-echo $$ > "$SST_PID"
 
 # give some time for stunnel from the previous SST to complete:
 check_round=0
@@ -293,26 +278,21 @@ while check_pid "$STUNNEL_PID" 1 "$STUNNEL_CONF"; do
     wsrep_log_info "Lingering stunnel daemon found at startup," \
                    "waiting for it to exit"
     check_round=$(( check_round+1 ))
-    if [ $check_round -eq 10 ]; then
+    if [ $check_round -eq 30 ]; then
-        wsrep_log_error "stunnel daemon still running."
+        wsrep_log_error "stunnel daemon still running..."
         exit 114 # EALREADY
     fi
     sleep 1
 done
 
-MODULE="${WSREP_SST_OPT_MODULE:-rsync_sst}"
-
-RSYNC_PID="$DATA/$MODULE.pid"
-RSYNC_CONF="$DATA/$MODULE.conf"
-
 # give some time for rsync from the previous SST to complete:
 check_round=0
 while check_pid "$RSYNC_PID" 1 "$RSYNC_CONF"; do
     wsrep_log_info "Lingering rsync daemon found at startup," \
                    "waiting for it to exit"
     check_round=$(( check_round+1 ))
-    if [ $check_round -eq 10 ]; then
+    if [ $check_round -eq 30 ]; then
-        wsrep_log_error "rsync daemon still running."
+        wsrep_log_error "rsync daemon still running..."
         exit 114 # EALREADY
     fi
     sleep 1