This is an automated email from the ASF dual-hosted git repository.

chesnay pushed a commit to branch master
in repository https://gitbox.apache.org/repos/asf/flink.git


The following commit(s) were added to refs/heads/master by this push:
     new 3436381  [FLINK-20659][yarn][tests] Add flink-yarn-test README
3436381 is described below

commit 3436381117ef8fb21f7e02ef63d6d6ec07792203
Author: Matthias Pohl <[email protected]>
AuthorDate: Wed Feb 24 18:15:32 2021 +0100

    [FLINK-20659][yarn][tests] Add flink-yarn-test README
---
 flink-yarn-tests/README.md                              | 17 +++++++++++++++++
 .../flink/yarn/YARNSessionCapacitySchedulerITCase.java  | 14 ++++++++++++++
 2 files changed, 31 insertions(+)

diff --git a/flink-yarn-tests/README.md b/flink-yarn-tests/README.md
new file mode 100644
index 0000000..998b03d
--- /dev/null
+++ b/flink-yarn-tests/README.md
@@ -0,0 +1,17 @@
+# Flink YARN tests
+
+`flink-yarn-test` collects test cases which are deployed to a local Apache 
Hadoop YARN cluster. 
+There are several things to consider when running these tests locally:
+
+* `YarnTestBase` spins up a `MiniYARNCluster`. This cluster spawns processes 
outside of the IDE's JVM 
+  to run the workers on. `JAVA_HOME` needs to be set to make this work.
+* The Flink cluster within each test is deployed using the `flink-dist` 
binaries. Any changes made 
+  to the code will only take effect after rebuilding the `flink-dist` module.
+* Each `YARN*ITCase` will have a local working directory for resources like 
logs to be stored. These 
+  working directories are located in `flink-yarn-tests/target/` (see 
+  `find flink-yarn-tests/target -name "*.err" -or -name "*.out"` for the 
test's output).
+* There is a known problem causing test instabilities due to our usage of 
Hadoop 2.8.3 executing the 
+  tests. This is caused by a bug 
[YARN-7007](https://issues.apache.org/jira/browse/YARN-7007) that 
+  got fixed in [Hadoop 
2.8.6](https://issues.apache.org/jira/projects/YARN/versions/12344056). See 
+  [FLINK-15534](https://issues.apache.org/jira/browse/FLINK-15534) for further 
details on the 
+  related discussion.
diff --git 
a/flink-yarn-tests/src/test/java/org/apache/flink/yarn/YARNSessionCapacitySchedulerITCase.java
 
b/flink-yarn-tests/src/test/java/org/apache/flink/yarn/YARNSessionCapacitySchedulerITCase.java
index 1f7d226..b8b328a 100644
--- 
a/flink-yarn-tests/src/test/java/org/apache/flink/yarn/YARNSessionCapacitySchedulerITCase.java
+++ 
b/flink-yarn-tests/src/test/java/org/apache/flink/yarn/YARNSessionCapacitySchedulerITCase.java
@@ -232,6 +232,10 @@ public class YARNSessionCapacitySchedulerITCase extends 
YarnTestBase {
      * <p>This ensures that with (any) pre-allocated off-heap memory by us, 
there is some off-heap
      * memory remaining for Flink's libraries. Creating task managers will 
thus fail if no off-heap
      * memory remains.
+     *
+     * @throws NullPointerException There is a known Hadoop bug (YARN-7007) 
that got fixed in Hadoop
+     *     2.8.6 but might cause test instabilities. See 
FLINK-20659/FLINK-15534 for further
+     *     information.
      */
     @Test
     public void perJobYarnClusterOffHeap() throws Exception {
@@ -289,6 +293,10 @@ public class YARNSessionCapacitySchedulerITCase extends 
YarnTestBase {
      *
      * <p><b>Hint: </b> If you think it is a good idea to add more assertions 
to this test, think
      * again!
+     *
+     * @throws NullPointerException There is a known Hadoop bug (YARN-7007) 
that got fixed in Hadoop
+     *     2.8.6 but might cause test instabilities. See 
FLINK-13009/FLINK-15534 for further
+     *     information.
      */
     @Test
     public void
@@ -451,6 +459,9 @@ public class YARNSessionCapacitySchedulerITCase extends 
YarnTestBase {
      * Test deployment to non-existing queue & ensure that the system logs a 
WARN message for the
      * user. (Users had unexpected behavior of Flink on YARN because they 
mistyped the target queue.
      * With an error message, we can help users identifying the issue)
+     *
+     * @throws NullPointerException There is a known Hadoop bug (YARN-7007) 
that got fixed in Hadoop
+     *     2.8.6 but might cause test instabilities. See FLINK-15534 for 
further information.
      */
     @Test
     public void testNonexistingQueueWARNmessage() throws Exception {
@@ -493,6 +504,9 @@ public class YARNSessionCapacitySchedulerITCase extends 
YarnTestBase {
     /**
      * Test per-job yarn cluster with the parallelism set at the CliFrontend 
instead of the YARN
      * client.
+     *
+     * @throws NullPointerException There is a known Hadoop bug (YARN-7007) 
that got fixed in Hadoop
+     *     2.8.6 but might cause test instabilities. See FLINK-15534 for 
further information.
      */
     @Test
     public void perJobYarnClusterWithParallelism() throws Exception {

Reply via email to