Github user NicoK commented on a diff in the pull request:
https://github.com/apache/flink/pull/4358#discussion_r139926257
--- Diff:
flink-fs-tests/src/test/java/org/apache/flink/hdfstests/HDFSTest.java ---
@@ -257,7 +265,88 @@ public void testBlobServerRecovery() throws Exception {
try {
blobStoreService =
BlobUtils.createBlobStoreFromConfig(config);
- BlobRecoveryITCase.testBlobServerRecovery(config,
blobStoreService);
+ BlobServerRecoveryTest.testBlobServerRecovery(config,
blobStoreService);
+ } finally {
+ if (blobStoreService != null) {
+ blobStoreService.closeAndCleanupAllData();
+ }
+ }
+ }
+
+ /**
+ * Tests that with {@link HighAvailabilityMode#ZOOKEEPER} distributed
corrupted JARs are
+ * recognised during the download via a {@link
org.apache.flink.runtime.blob.BlobServer}.
+ */
+ @Test
+ public void testBlobServerCorruptedFile() throws Exception {
+ org.apache.flink.configuration.Configuration
+ config = new
org.apache.flink.configuration.Configuration();
+ config.setString(HighAvailabilityOptions.HA_MODE, "ZOOKEEPER");
+ config.setString(CoreOptions.STATE_BACKEND, "ZOOKEEPER");
--- End diff --
I'm pretty sure, this should have been `FILESYSTEM` and I was just to eager
to change everything to `ZOOKEEPER` when copying the tests over... Anyway, this
option is not used at all because we only test the BLOB service classes (also
in the test classes I copied this from).
---
