Moved cudf to 0.14 for CI (NVIDIA#152)

nartal1 · Jun 11, 2020 · dafcd4e · dafcd4e
1 parent 915fdf7
commit dafcd4e
Show file tree

Hide file tree

Showing 6 changed files with 9 additions and 6 deletions.
diff --git a/integration_tests/README.md b/integration_tests/README.md
@@ -39,7 +39,7 @@ Most clusters probably will not have the RAPIDS plugin installed in the cluster
 If just want to verify the SQL replacement is working you will need to add the `rapids-4-spark` and `cudf` jars to your `spark-submit` command.
 
 ```
-$SPARK_HOME/bin/spark-submit --jars "rapids-4-spark_2.12-0.1-SNAPSHOT.jar,cudf-0.14-SNAPSHOT.jar" ./runtests.py
+$SPARK_HOME/bin/spark-submit --jars "rapids-4-spark_2.12-0.1-SNAPSHOT.jar,cudf-0.14.jar" ./runtests.py
 ```
 
 You don't have to enable the plugin for this to work, the test framework will do that for you.
@@ -70,7 +70,7 @@ The TPCxBB, TPCH, and Mortgage tests in this framework can be enabled by providi
 As an example, here is the `spark-submit` command with the TPCxBB parameters:
 
 ```
-$SPARK_HOME/bin/spark-submit --jars "rapids-4-spark_2.12-0.1-SNAPSHOT.jar,cudf-0.14-SNAPSHOT.jar,rapids-4-spark-tests_2.12-0.1-SNAPSHOT.jar" ./runtests.py --tpcxbb_format="csv" --tpcxbb_path="/path/to/tpcxbb/csv"
+$SPARK_HOME/bin/spark-submit --jars "rapids-4-spark_2.12-0.1-SNAPSHOT.jar,cudf-0.14.jar,rapids-4-spark-tests_2.12-0.1-SNAPSHOT.jar" ./runtests.py --tpcxbb_format="csv" --tpcxbb_path="/path/to/tpcxbb/csv"
 ```
 
 ## Writing tests

diff --git a/jenkins/Jenkinsfile.integration b/jenkins/Jenkinsfile.integration
@@ -32,7 +32,7 @@ pipeline {
     }
 
     parameters {
-        string(name: 'CUDF_VER', defaultValue: '0.14-SNAPSHOT',
+        string(name: 'CUDF_VER', defaultValue: '0.14',
             description: '-Dcudf.version= \n\n Default for cudf version')
         string(name: 'CUDA_CLASSIFIER', defaultValue: '',
             description: '-Dclassifier=\n\n cuda10-1, cuda10-2, EMPTY as cuda10-1')

diff --git a/jenkins/Jenkinsfile.nightly b/jenkins/Jenkinsfile.nightly
@@ -63,7 +63,7 @@ pipeline {
                         -v ${HOME}/.zinc:${HOME}/.zinc:rw \
                         -v /etc/passwd:/etc/passwd -v /etc/group:/etc/group") {
                         sh "mvn -U -B clean deploy $MVN_URM_MIRROR"
-                        sh "jenkins/printJarVersion.sh 'CUDFVersion' '${HOME}/.m2/repository/ai/rapids/cudf/0.14-SNAPSHOT' 'cudf-0.14-' '-cuda10-1.jar'"
+                        sh "jenkins/printJarVersion.sh 'CUDFVersion' '${HOME}/.m2/repository/ai/rapids/cudf/0.14' 'cudf-0.14' '-cuda10-1.jar'"
                         sh "jenkins/printJarVersion.sh 'SPARKVersion' '${HOME}/.m2/repository/org/apache/spark/spark-core_2.12/3.0.1-SNAPSHOT' 'spark-core_2.12-3.0.1-' '.jar'"
                     }
                 }
@@ -78,7 +78,7 @@ pipeline {
                     build(job: 'spark/rapids_integration-0.1-github',
                           propagate: false,
                           parameters: [string(name: 'REF', value: 'branch-0.1'),
-                                       string(name: 'CUDF_VER', value: '0.14-SNAPSHOT'),
+                                       string(name: 'CUDF_VER', value: '0.14'),
                                        booleanParam(name: 'BUILD_CENTOS7', value: false),])
 
                     slack("#rapidsai-spark-cicd", "Success", color: "#33CC33")

diff --git a/jenkins/spark-premerge-build.sh b/jenkins/spark-premerge-build.sh
@@ -17,6 +17,8 @@
 
 set -ex
 
+nvidia-smi
+
 if [ "$SPARK_VER"x == x ];then
     SPARK_VER="3.0.1-SNAPSHOT"
 fi

diff --git a/jenkins/spark-tests.sh b/jenkins/spark-tests.sh
@@ -17,7 +17,7 @@
 
 set -ex
 if [ "$CUDF_VER"x == x ];then
-    CUDF_VER="0.14-SNAPSHOT"
+    CUDF_VER="0.14"
 fi
 
 if [ "$PROJECT_VER"x == x ];then

diff --git a/sql-plugin/src/main/scala/ai/rapids/spark/GpuDeviceManager.scala b/sql-plugin/src/main/scala/ai/rapids/spark/GpuDeviceManager.scala
@@ -57,6 +57,7 @@ object GpuDeviceManager extends Logging {
       GpuDeviceManager.setGpuDeviceAndAcquire(addr)
     } catch {
       case NonFatal(e) =>
+        logInfo(s"Will not use GPU $addr because of $e")
         // we may have lost a race trying to acquire this addr or GPU is already busy
         return false
     }