MDEV-28423: Galera IST is failing on Joiner node

This commit fixes an issue with IST handling in
version 10.9 which is a regression after MDEV-26971
and related to trying to get a non-existent "total"
tag on the IST branch (this tag is only defined in
SST mode).

2 files changed, 28 insertions, 27 deletions

diff --git a/scripts/wsrep_sst_mariabackup.sh b/scripts/wsrep_sst_mariabackup.sh
index 87ed8392475..314d2349bcc 100644
--- a/scripts/wsrep_sst_mariabackup.sh
+++ b/scripts/wsrep_sst_mariabackup.sh
@@ -86,6 +86,7 @@ encrypt_threads=""
 encrypt_chunk=""
 
 readonly SECRET_TAG='secret'
+readonly TOTAL_TAG='total'
 
 # Required for backup locks
 # For backup locks it is 1 sent by joiner
@@ -324,7 +325,8 @@ get_transfer()
             if [ -z "$ssl_dhparams" ]; then
                 # Determine the socat version
                 SOCAT_VERSION=$(socat -V 2>&1 | \
-                                grep -m1 -owE '[0-9]+(\.[0-9]+)+' | head -n1)
+                                grep -m1 -owE '[0-9]+(\.[0-9]+)+' | \
+                                head -n1 || :)
                 if [ -z "$SOCAT_VERSION" ]; then
                     wsrep_log_error "******** FATAL ERROR ******************"
                     wsrep_log_error "* Cannot determine the socat version. *"
@@ -770,25 +772,26 @@ recv_joiner()
 
     if [ $checkf -eq 1 ]; then
         if [ ! -r "$MAGIC_FILE" ]; then
-            # this message should cause joiner to abort
+            # this message should cause joiner to abort:
             wsrep_log_error "receiving process ended without creating" \
-                            "'$MAGIC_FILE'"
-            wsrep_log_info "Contents of datadir"
+                            "magic file ($MAGIC_FILE)"
+            wsrep_log_info "Contents of datadir:"
             wsrep_log_info $(ls -l "$dir/"*)
             exit 32
         fi
 
         # check donor supplied secret
-        SECRET=$(grep -F -- "$SECRET_TAG " "$MAGIC_FILE" 2>/dev/null | \
-                 cut -d ' ' -f2)
+        SECRET=$(grep -m1 -E "^$SECRET_TAG[[:space:]]" -- "$MAGIC_FILE" || :)
+        SECRET=$(trim_string "${SECRET#$SECRET_TAG}")
         if [ "$SECRET" != "$MY_SECRET" ]; then
             wsrep_log_error "Donor does not know my secret!"
             wsrep_log_info "Donor: '$SECRET', my: '$MY_SECRET'"
             exit 32
         fi
 
-        # remove secret from the magic file
-        grep -v -F -- "$SECRET_TAG " "$MAGIC_FILE" > "$MAGIC_FILE.new"
+        # remove secret and total from the magic file
+        grep -v -E "^($SECRET_TAG|$TOTAL_TAG)[[:space:]]" -- \
+             "$MAGIC_FILE" > "$MAGIC_FILE.new"
         mv "$MAGIC_FILE.new" "$MAGIC_FILE"
     fi
 }
@@ -1256,8 +1259,7 @@ then
 
     recv_joiner "$STATDIR" "$stagemsg-gtid" $stimeout 1 1
 
-    if ! ps -p "$WSREP_SST_OPT_PARENT" >/dev/null 2>&1
-    then
+    if ! ps -p "$WSREP_SST_OPT_PARENT" >/dev/null 2>&1; then
         wsrep_log_error "Parent mysqld process (PID: $WSREP_SST_OPT_PARENT)" \
                         "terminated unexpectedly."
         exit 32
diff --git a/scripts/wsrep_sst_rsync.sh b/scripts/wsrep_sst_rsync.sh
index 29e2b390e27..27d4e875674 100644
--- a/scripts/wsrep_sst_rsync.sh
+++ b/scripts/wsrep_sst_rsync.sh
@@ -804,8 +804,7 @@ EOF
         sleep 1
     done
 
-    if ! ps -p $MYSQLD_PID >/dev/null 2>&1
-    then
+    if ! ps -p $MYSQLD_PID >/dev/null 2>&1; then
         wsrep_log_error \
             "Parent mysqld process (PID: $MYSQLD_PID) terminated unexpectedly."
         kill -- -$MYSQLD_PID
@@ -813,24 +812,24 @@ EOF
         exit 32
     fi
 
-    if [ -r "$MAGIC_FILE" ]; then
-        if [ -n "$MY_SECRET" ]; then
-            # Check donor supplied secret:
-            SECRET=$(grep -F -- "$SECRET_TAG " "$MAGIC_FILE" 2>/dev/null | \
-                     cut -d ' ' -f2)
-            if [ "$SECRET" != "$MY_SECRET" ]; then
-                wsrep_log_error "Donor does not know my secret!"
-                wsrep_log_info "Donor: '$SECRET', my: '$MY_SECRET'"
-                exit 32
-            fi
-        fi
-    else
+    if [ ! -r "$MAGIC_FILE" ]; then
         # This message should cause joiner to abort:
-        wsrep_log_info "rsync process ended without creating magic file"
-        echo "rsync process ended without creating '$MAGIC_FILE'"
+        wsrep_log_info "rsync process ended without creating" \
+                       "magic file ($MAGIC_FILE)"
         exit 32
     fi
 
+    if [ -n "$MY_SECRET" ]; then
+        # Check donor supplied secret:
+        SECRET=$(grep -m1 -E "^$SECRET_TAG[[:space:]]" -- "$MAGIC_FILE" || :)
+        SECRET=$(trim_string "${SECRET#$SECRET_TAG}")
+        if [ "$SECRET" != "$MY_SECRET" ]; then
+            wsrep_log_error "Donor does not know my secret!"
+            wsrep_log_info "Donor: '$SECRET', my: '$MY_SECRET'"
+            exit 32
+        fi
+    fi
+
     if [ -n "$WSREP_SST_OPT_BINLOG" ]; then
         binlog_tar_present=0
         [ -f "$BINLOG_TAR_FILE" ] && binlog_tar_present=1
@@ -907,7 +906,7 @@ EOF
     if [ -n "$MY_SECRET" ]; then
         # remove secret from the magic file, and output
         # the UUID:seqno & wsrep_gtid_domain_id:
-        grep -v -F -- "$SECRET_TAG " "$MAGIC_FILE"
+        grep -v -E "^$SECRET_TAG[[:space:]]" -- "$MAGIC_FILE"
     else
         # Output the UUID:seqno and wsrep_gtid_domain_id:
         cat "$MAGIC_FILE"