diff --git a/sql/catalyst/src/main/scala/org/apache/spark/sql/internal/SQLConf.scala b/sql/catalyst/src/main/scala/org/apache/spark/sql/internal/SQLConf.scala
index 4f77c54a7af5..80cdc61484c0 100644
--- a/sql/catalyst/src/main/scala/org/apache/spark/sql/internal/SQLConf.scala
+++ b/sql/catalyst/src/main/scala/org/apache/spark/sql/internal/SQLConf.scala
@@ -27,11 +27,13 @@ import scala.util.matching.Regex
 
 import org.apache.hadoop.fs.Path
 
+import org.apache.spark.{SparkContext, SparkEnv}
 import org.apache.spark.internal.Logging
 import org.apache.spark.internal.config._
 import org.apache.spark.network.util.ByteUnit
 import org.apache.spark.sql.catalyst.analysis.Resolver
 import org.apache.spark.sql.catalyst.expressions.codegen.CodeGenerator
+import org.apache.spark.util.Utils
 
 ////////////////////////////////////////////////////////////////////////////////////////////////////
 // This file defines the configuration options for Spark SQL.
@@ -70,7 +72,7 @@ object SQLConf {
    * Default config. Only used when there is no active SparkSession for the thread.
    * See [[get]] for more information.
    */
-  private val fallbackConf = new ThreadLocal[SQLConf] {
+  private lazy val fallbackConf = new ThreadLocal[SQLConf] {
     override def initialValue: SQLConf = new SQLConf
   }
 
@@ -1087,6 +1089,12 @@ object SQLConf {
 class SQLConf extends Serializable with Logging {
   import SQLConf._
 
+  if (Utils.isTesting && SparkEnv.get != null) {
+    // assert that we're only accessing it on the driver.
+    assert(SparkEnv.get.executorId == SparkContext.DRIVER_IDENTIFIER,
+      "SQLConf should only be created and accessed on the driver.")
+  }
+
   /** Only low degree of contention is expected for conf, thus NOT using ConcurrentHashMap. */
   @transient protected[spark] val settings = java.util.Collections.synchronizedMap(
     new java.util.HashMap[String, String]())