feat(mariadb): skip user-input SQL errors in reconfigure action to avoid infinite retry (#2889)

Author: weicao

addons/mariadb/scripts-ut-spec/reconfigure_persisted_alpha86_spec.sh

@@ -538,6 +538,27 @@ Describe "alpha.86 reconfigureAction.persisted semisync gates"
     End
   End
 
+  Describe "Rejected user-input SQL errors do not keep reconfigure on the failing path"
+    It "base reconfigureAction skips classified engine rejects without exiting 1"
+      When call extract_base_helper_body
+      The status should be success
+      The output should include 'skipped_count=$((skipped_count + 1))'
+      The output should include 'parameter(s) were rejected by engine and skipped'
+      The output should include 'if [ "${applied_count}" -eq 0 ] && [ "${skipped_count}" -eq 0 ]; then'
+      The output should not include 'failing reconfigure to avoid accepting invalid rendered config'
+    End
+
+    It "persisted reconfigureAction skips classified engine rejects without writing overrides or exiting 1"
+      When call extract_persisted_helper_body
+      The status should be success
+      The output should include 'skipped_count=$((skipped_count + 1))'
+      The output should include 'continue'
+      The output should include 'parameter(s) were rejected by engine and skipped'
+      The output should include 'if [ "${applied_count}" -eq 0 ] && [ "${skipped_count}" -eq 0 ]; then'
+      The output should not include 'failing reconfigure to avoid accepting invalid rendered config'
+    End
+  End
+
   Describe "Regression: KB-managed config has NO !includedir (alpha.84 v1 parser FAIL)"
     It "config/mariadb-semisync.tpl does NOT contain !includedir (mariadbd loads via --defaults-extra-file instead)"
       # alpha.84 v1 put !includedir into KB-managed my.cnf, which KB

addons/mariadb/templates/_helpers.tpl

@@ -428,6 +428,7 @@ reconfigure:
         fi
 
         applied_count=0
+        skipped_count=0
         while IFS= read -r assignment; do
           [ -n "${assignment}" ] || continue
           case "${assignment}" in
@@ -453,12 +454,28 @@ reconfigure:
             echo "Set parameter ${param_name} to value ${param_value}"
             applied_count=$((applied_count + 1))
           else
-            echo "Failed to set parameter ${param_name}=${param_value}: ${output}" >&2
-            exit 1
+            # Classify MariaDB SQL errors that are caused by bad user input.
+            # These should not be retried indefinitely by the controller.
+            error_code=$(printf '%s' "${output}" | grep -oE 'ERROR [0-9]+' | head -1 | awk '{print $2}')
+            case "${error_code}" in
+              1231|1232|1193|1064)
+                echo "[REJECT] parameter ${param_name}=${param_value} rejected by engine (error ${error_code}): ${output}"
+                echo "[REJECT] parameter ${param_name}=${param_value} rejected by engine (error ${error_code}): ${output}" >&2
+                skipped_count=$((skipped_count + 1))
+                ;;
+              *)
+                echo "Failed to set parameter ${param_name}=${param_value}: ${output}" >&2
+                exit 1
+                ;;
+            esac
           fi
         done < "${parameter_file}"
 
-        if [ "${applied_count}" -eq 0 ]; then
+        if [ "${skipped_count}" -gt 0 ]; then
+          echo "${skipped_count} parameter(s) were rejected by engine and skipped"
+          echo "${skipped_count} parameter(s) were rejected by engine and skipped" >&2
+        fi
+        if [ "${applied_count}" -eq 0 ] && [ "${skipped_count}" -eq 0 ]; then
           echo "No parameters were applied during reconfigure action" >&2
           exit 1
         fi
@@ -846,6 +863,7 @@ reconfigure:
         }
 
         applied_count=0
+        skipped_count=0
         while IFS= read -r assignment; do
           [ -n "${assignment}" ] || continue
           case "${assignment}" in
@@ -892,8 +910,20 @@ reconfigure:
             echo "Set parameter ${param_name} to value ${param_value}"
             applied_count=$((applied_count + 1))
           else
-            echo "Failed to set parameter ${param_name}=${param_value}: ${output}" >&2
-            exit 1
+            # Classify MariaDB SQL errors that are caused by bad user input.
+            error_code=$(printf '%s' "${output}" | grep -oE 'ERROR [0-9]+' | head -1 | awk '{print $2}')
+            case "${error_code}" in
+              1231|1232|1193|1064)
+                echo "[REJECT] parameter ${param_name}=${param_value} rejected by engine (error ${error_code}): ${output}"
+                echo "[REJECT] parameter ${param_name}=${param_value} rejected by engine (error ${error_code}): ${output}" >&2
+                skipped_count=$((skipped_count + 1))
+                continue
+                ;;
+              *)
+                echo "Failed to set parameter ${param_name}=${param_value}: ${output}" >&2
+                exit 1
+                ;;
+            esac
           fi
 
           # alpha.86 v1 — persist the override AFTER SET GLOBAL succeeds.
@@ -956,7 +986,11 @@ reconfigure:
           #     than catching the same issue at write time.
         done < "${parameter_file}"
 
-        if [ "${applied_count}" -eq 0 ]; then
+        if [ "${skipped_count}" -gt 0 ]; then
+          echo "${skipped_count} parameter(s) were rejected by engine and skipped"
+          echo "${skipped_count} parameter(s) were rejected by engine and skipped" >&2
+        fi
+        if [ "${applied_count}" -eq 0 ] && [ "${skipped_count}" -eq 0 ]; then
           echo "No parameters were applied during reconfigure action" >&2
           exit 1
         fi