Apache
/
hadoop
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

HDFS-15204. TestRetryCacheWithHA testRemoveCacheDescriptor fails intermittently. Contributed by Ahmed Hussein. 

(cherry picked from commit 1d4d0fcbe1)
This commit is contained in:
Inigo Goiri 2020-03-04 10:02:54 -08:00
parent 369f4f9c58
commit 1d89ba72f6
1 changed files with 39 additions and 32 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

59
hadoop-hdfs-project/hadoop-hdfs/src/test/java/org/apache/hadoop/hdfs/server/namenode/ha/TestRetryCacheWithHA.java
View File

@ -34,8 +34,10 @@ import java.util.Iterator;
import java.util.Map;
import java.util.Random;
import java.util.Set;
import java.util.concurrent.ConcurrentHashMap;
import java.util.concurrent.atomic.AtomicBoolean;
import org.apache.hadoop.test.GenericTestUtils;
import org.slf4j.Logger;
import org.slf4j.LoggerFactory;
import org.apache.hadoop.conf.Configuration;
@ -1296,7 +1298,7 @@ public class TestRetryCacheWithHA {
*/
public void testClientRetryWithFailover(final AtMostOnceOp op)
throws Exception {
final Map<String, Object> results = new HashMap<String, Object>();
final Map<String, Object> results = new ConcurrentHashMap<>();
op.prepare();
// set DummyRetryInvocationHandler#block to true
@ -1309,10 +1311,7 @@ public class TestRetryCacheWithHA {
op.invoke();
Object result = op.getResult();
LOG.info("Operation " + op.name + " finished");
synchronized (TestRetryCacheWithHA.this) {
results.put(op.name, result == null ? "SUCCESS" : result);
TestRetryCacheWithHA.this.notifyAll();
}
} catch (Exception e) {
LOG.info("Got Exception while calling " + op.name, e);
} finally {
@ -1333,39 +1332,47 @@ public class TestRetryCacheWithHA {
LOG.info("Setting block to false");
DummyRetryInvocationHandler.block.set(false);
synchronized (this) {
while (!results.containsKey(op.name)) {
this.wait();
}
GenericTestUtils.waitFor(() -> results.containsKey(op.name), 5, 10000);
LOG.info("Got the result of " + op.name + ": "
+ results.get(op.name));
}
// Waiting for failover.
while (cluster.getNamesystem(1).isInStandbyState()) {
Thread.sleep(10);
}
GenericTestUtils
.waitFor(() -> !cluster.getNamesystem(1).isInStandbyState(), 5, 10000);
long hitNN0 = cluster.getNamesystem(0).getRetryCache().getMetricsForTests()
final long[] hitsNN = new long[]{0, 0};
GenericTestUtils.waitFor(() -> {
hitsNN[0] = cluster.getNamesystem(0).getRetryCache()
.getMetricsForTests()
.getCacheHit();
long hitNN1 = cluster.getNamesystem(1).getRetryCache().getMetricsForTests()
hitsNN[1] = cluster.getNamesystem(1).getRetryCache()
.getMetricsForTests()
.getCacheHit();
assertTrue("CacheHit: " + hitNN0 + ", " + hitNN1,
hitNN0 + hitNN1 > 0);
long updatedNN0 = cluster.getNamesystem(0).getRetryCache()
.getMetricsForTests().getCacheUpdated();
long updatedNN1 = cluster.getNamesystem(1).getRetryCache()
.getMetricsForTests().getCacheUpdated();
return (hitsNN[0] + hitsNN[1]) > 0;
}, 5, 10000);
assertTrue("CacheHit: " + hitsNN[0] + ", " + hitsNN[1],
+hitsNN[0] + hitsNN[1] > 0);
final long[] updatesNN = new long[]{0, 0};
GenericTestUtils.waitFor(() -> {
updatesNN[0] = cluster.getNamesystem(0).getRetryCache()
.getMetricsForTests()
.getCacheUpdated();
updatesNN[1] = cluster.getNamesystem(1).getRetryCache()
.getMetricsForTests()
.getCacheUpdated();
return updatesNN[0] > 0 && updatesNN[1] > 0;
}, 5, 10000);
// Cache updated metrics on NN0 should be >0 since the op was process on NN0
assertTrue("CacheUpdated on NN0: " + updatedNN0, updatedNN0 > 0);
assertTrue("CacheUpdated on NN0: " + updatesNN[0], updatesNN[0] > 0);
// Cache updated metrics on NN0 should be >0 since NN1 applied the editlog
assertTrue("CacheUpdated on NN1: " + updatedNN1, updatedNN1 > 0);
assertTrue("CacheUpdated on NN1: " + updatesNN[1], updatesNN[1] > 0);
long expectedUpdateCount = op.getExpectedCacheUpdateCount();
if (expectedUpdateCount > 0) {
assertEquals("CacheUpdated on NN0: " + updatedNN0, expectedUpdateCount,
updatedNN0);
assertEquals("CacheUpdated on NN0: " + updatedNN1, expectedUpdateCount,
updatedNN1);
assertEquals("CacheUpdated on NN0: " + updatesNN[0], expectedUpdateCount,
updatesNN[0]);
assertEquals("CacheUpdated on NN0: " + updatesNN[1], expectedUpdateCount,
updatesNN[1]);
}
}