Fix allowMultipleJars recommend setting message (#9922)

* Fix multiple jar recommend setting message

Signed-off-by: Haoyang Li <haoyangl@nvidia.com>

* verify

Signed-off-by: Haoyang Li <haoyangl@nvidia.com>

* Explain why multiple jars is bad

Signed-off-by: Haoyang Li <haoyangl@nvidia.com>

* Always is bad, and also warning it in require

Signed-off-by: Haoyang Li <haoyangl@nvidia.com>

---------

Signed-off-by: Haoyang Li <haoyangl@nvidia.com>

docs/additional-functionality/advanced_configs.md

@@ -58,6 +58,7 @@ Name | Description | Default Value | Applicable at
 <a name="shuffle.ucx.activeMessages.forceRndv"></a>spark.rapids.shuffle.ucx.activeMessages.forceRndv|Set to true to force 'rndv' mode for all UCX Active Messages. This should only be required with UCX 1.10.x. UCX 1.11.x deployments should set to false.|false|Startup
 <a name="shuffle.ucx.managementServerHost"></a>spark.rapids.shuffle.ucx.managementServerHost|The host to be used to start the management server|null|Startup
 <a name="shuffle.ucx.useWakeup"></a>spark.rapids.shuffle.ucx.useWakeup|When set to true, use UCX's event-based progress (epoll) in order to wake up the progress thread when needed, instead of a hot loop.|true|Startup
+<a name="sql.allowMultipleJars"></a>spark.rapids.sql.allowMultipleJars|Allow multiple rapids-4-spark, spark-rapids-jni, and cudf jars on the classpath. Spark will take the first one it finds, so the version may not be expected. Possisble values are ALWAYS: allow all jars, SAME_REVISION: only allow jars with the same revision, NEVER: do not allow multiple jars at all.|SAME_REVISION|Startup
 <a name="sql.castDecimalToFloat.enabled"></a>spark.rapids.sql.castDecimalToFloat.enabled|Casting from decimal to floating point types on the GPU returns results that have tiny difference compared to results returned from CPU.|true|Runtime
 <a name="sql.castFloatToDecimal.enabled"></a>spark.rapids.sql.castFloatToDecimal.enabled|Casting from floating point types to decimal on the GPU returns results that have tiny difference compared to results returned from CPU.|true|Runtime
 <a name="sql.castFloatToIntegralTypes.enabled"></a>spark.rapids.sql.castFloatToIntegralTypes.enabled|Casting from floating point types to integral types on the GPU supports a slightly different range of values when using Spark 3.1.0 or later. Refer to the CAST documentation for more details.|true|Runtime

sql-plugin/src/main/scala/com/nvidia/spark/rapids/Plugin.scala

@@ -147,26 +147,30 @@ object RapidsPluginUtils extends Logging {
         }.mkString + "\n"
       }
     }.mkString
-    // scalastyle:off line.size.limit
-    lazy val msg = s"""Multiple $jarName jars found in the classpath:
-        |$rapidsJarsVersMsg
-        |Please make sure there is only one $jarName jar in the classpath.
-        |If it is impossible to fix the classpath you can suppress the error by setting ${RapidsConf.ALLOW_MULTIPLE_JARS.key} to SAME_REVISION or ALWAYS.
-        """.stripMargin
-    // scalastyle:on line.size.limit
+    lazy val msg = s"Multiple $jarName jars found in the classpath:\n$rapidsJarsVersMsg" +
+        s"Please make sure there is only one $jarName jar in the classpath. "
+
+    require(revisionMap.size > 0, s"Could not find any $jarName jars in the classpath")
 
     conf.allowMultipleJars match {
       case AllowMultipleJars.ALWAYS =>
         if (revisionMap.size != 1 || revisionMap.values.exists(_.size != 1)) {
           logWarning(msg)
         }
       case AllowMultipleJars.SAME_REVISION =>
-        require(revisionMap.size == 1, msg)
+        val recommended = "If it is impossible to fix the classpath you can suppress the " +
+              s"error by setting ${RapidsConf.ALLOW_MULTIPLE_JARS.key} to ALWAYS, but this " +
+              s"can cause unpredictable behavior as the plugin may pick up the wrong jar."
+        require(revisionMap.size == 1, msg + recommended)
         if (revisionMap.values.exists(_.size != 1)) {
-          logWarning(msg)
+          logWarning(msg + recommended)
         }
       case AllowMultipleJars.NEVER =>
-        require(revisionMap.size == 1 && revisionMap.values.forall(_.size == 1), msg)
+        val recommended = "If it is impossible to fix the classpath you can suppress the " +
+            s"error by setting ${RapidsConf.ALLOW_MULTIPLE_JARS.key} to SAME_REVISION or ALWAYS." +
+            " But setting it to ALWAYS can cause unpredictable behavior as the plugin may pick " +
+            "up the wrong jar."
+        require(revisionMap.size == 1 && revisionMap.values.forall(_.size == 1), msg + recommended)
     }
   }
 

sql-plugin/src/main/scala/com/nvidia/spark/rapids/RapidsConf.scala

@@ -1845,7 +1845,6 @@ object RapidsConf {
   }
 
   val ALLOW_MULTIPLE_JARS = conf("spark.rapids.sql.allowMultipleJars")
-    .internal()
     .startupOnly()
     .doc("Allow multiple rapids-4-spark, spark-rapids-jni, and cudf jars on the classpath. " +
       "Spark will take the first one it finds, so the version may not be expected. Possisble " +