big-data/hadoop
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

HDFS-14993. checkDiskError doesn't work during datanode startup. Contributed by Yang Yun.

Browse Source
This commit is contained in:
Ayush Saxena 2020-01-28 22:59:26 +05:30
parent 08251538fe
commit 87c198468b
2 changed files with 34 additions and 3 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

6
hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/server/datanode/DataNode.java
View File

@ -1701,14 +1701,14 @@ void initBlockPool(BPOfferService bpos) throws IOException {
// the dataset, block scanners, etc.
initStorage(nsInfo);
// Exclude failed disks before initializing the block pools to avoid startup
// failures.
checkDiskError();
try {
data.addBlockPool(nsInfo.getBlockPoolID(), getConf());
} catch (AddBlockPoolException e) {
handleAddBlockPoolError(e);
}
// HDFS-14993: check disk after add the block pool info.
checkDiskError();
blockScanner.enableBlockPoolId(bpos.getBlockPoolId());
initDirectoryScanner(getConf());
initDiskBalancer(data, getConf());

31
hadoop-hdfs-project/hadoop-hdfs/src/test/java/org/apache/hadoop/hdfs/server/datanode/TestDataNodeVolumeFailure.java
View File

@ -916,4 +916,35 @@ public boolean isSimulated() {
}
}
}
/*
* Verify the failed volume can be cheched during dn startup
*/
@Test(timeout = 120000)
public void testVolumeFailureDuringStartup() throws Exception {
LOG.debug("Data dir: is " + dataDir.getPath());
// fail the volume
data_fail = cluster.getInstanceStorageDir(1, 0);
failedDir = MiniDFSCluster.getFinalizedDir(data_fail,
cluster.getNamesystem().getBlockPoolId());
failedDir.setReadOnly();
// restart the dn
cluster.restartDataNode(1);
final DataNode dn = cluster.getDataNodes().get(1);
// should get the failed volume during startup
GenericTestUtils.waitFor(new Supplier<Boolean>() {
@Override
public Boolean get() {
return dn.getFSDataset() !=null &&
dn.getFSDataset().getVolumeFailureSummary() != null &&
dn.getFSDataset().getVolumeFailureSummary().
getFailedStorageLocations()!= null &&
dn.getFSDataset().getVolumeFailureSummary().
getFailedStorageLocations().length == 1;
}
}, 10, 30 * 1000);
}
}