HDFS-15820. Ensure snapshot root trash provisioning happens only post safe mode exit (#2682)

This commit is contained in:
bshashikant 2021-02-06 21:56:12 +05:30 committed by GitHub
parent c4918fb298
commit 19ae0faacc
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
5 changed files with 58 additions and 30 deletions

View File

@ -425,7 +425,7 @@ boolean leaveSafeMode(boolean force) {
BlockManagerSafeMode.STEP_AWAITING_REPORTED_BLOCKS);
prog.endPhase(Phase.SAFEMODE);
}
namesystem.checkAndProvisionSnapshotTrashRoots();
return true;
}

View File

@ -124,7 +124,8 @@
import org.apache.hadoop.hdfs.server.namenode.metrics.ReplicatedBlocksMBean;
import org.apache.hadoop.hdfs.server.protocol.SlowDiskReports;
import org.apache.hadoop.ipc.ObserverRetryOnActiveException;
import org.apache.hadoop.util.Time;
import org.apache.hadoop.util.*;
import static org.apache.hadoop.util.Time.now;
import static org.apache.hadoop.util.Time.monotonicNow;
import static org.apache.hadoop.hdfs.server.namenode.top.metrics.TopMetrics.TOPMETRICS_METRICS_SOURCE_NAME;
@ -329,11 +330,6 @@
import org.apache.hadoop.security.token.Token;
import org.apache.hadoop.security.token.TokenIdentifier;
import org.apache.hadoop.security.token.delegation.DelegationKey;
import org.apache.hadoop.util.Daemon;
import org.apache.hadoop.util.DataChecksum;
import org.apache.hadoop.util.ReflectionUtils;
import org.apache.hadoop.util.StringUtils;
import org.apache.hadoop.util.VersionInfo;
import org.apache.log4j.Logger;
import org.apache.log4j.Appender;
import org.apache.log4j.AsyncAppender;
@ -8531,25 +8527,37 @@ void checkAccess(String src, FsAction mode) throws IOException {
* Check if snapshot roots are created for all existing snapshottable
* directories. Create them if not.
*/
void checkAndProvisionSnapshotTrashRoots() throws IOException {
SnapshottableDirectoryStatus[] dirStatusList = getSnapshottableDirListing();
if (dirStatusList == null) {
return;
}
for (SnapshottableDirectoryStatus dirStatus : dirStatusList) {
String currDir = dirStatus.getFullPath().toString();
if (!currDir.endsWith(Path.SEPARATOR)) {
currDir += Path.SEPARATOR;
}
String trashPath = currDir + FileSystem.TRASH_PREFIX;
HdfsFileStatus fileStatus = getFileInfo(trashPath, false, false, false);
if (fileStatus == null) {
LOG.info("Trash doesn't exist for snapshottable directory {}. "
+ "Creating trash at {}", currDir, trashPath);
PermissionStatus permissionStatus = new PermissionStatus(getRemoteUser()
.getShortUserName(), null, SHARED_TRASH_PERMISSION);
mkdirs(trashPath, permissionStatus, false);
@Override
public void checkAndProvisionSnapshotTrashRoots() {
if (isSnapshotTrashRootEnabled) {
try {
SnapshottableDirectoryStatus[] dirStatusList =
getSnapshottableDirListing();
if (dirStatusList == null) {
return;
}
for (SnapshottableDirectoryStatus dirStatus : dirStatusList) {
String currDir = dirStatus.getFullPath().toString();
if (!currDir.endsWith(Path.SEPARATOR)) {
currDir += Path.SEPARATOR;
}
String trashPath = currDir + FileSystem.TRASH_PREFIX;
HdfsFileStatus fileStatus = getFileInfo(trashPath, false, false, false);
if (fileStatus == null) {
LOG.info("Trash doesn't exist for snapshottable directory {}. " + "Creating trash at {}", currDir, trashPath);
PermissionStatus permissionStatus =
new PermissionStatus(getRemoteUser().getShortUserName(), null,
SHARED_TRASH_PERMISSION);
mkdirs(trashPath, permissionStatus, false);
}
}
} catch (IOException e) {
final String msg =
"Could not provision Trash directory for existing "
+ "snapshottable directories. Exiting Namenode.";
ExitUtil.terminate(1, msg);
}
}
}

View File

@ -2011,9 +2011,6 @@ public HAState getState() {
public void startActiveServices() throws IOException {
try {
namesystem.startActiveServices();
if (namesystem.isSnapshotTrashRootEnabled()) {
namesystem.checkAndProvisionSnapshotTrashRoots();
}
startTrashEmptier(getConf());
} catch (Throwable t) {
doImmediateShutdown(t);

View File

@ -27,7 +27,9 @@
/** Namesystem operations. */
@InterfaceAudience.Private
public interface Namesystem extends RwLock, SafeMode {
/** Is this name system running? */
/**
* Is this name system running?
*/
boolean isRunning();
BlockCollection getBlockCollection(long id);
@ -55,4 +57,10 @@ public interface Namesystem extends RwLock, SafeMode {
* @throws IOException
*/
void removeXattr(long id, String xattrName) throws IOException;
/**
* Check if snapshot roots are created for all existing snapshottable
* directories. Create them if not.
*/
void checkAndProvisionSnapshotTrashRoots();
}

View File

@ -2524,7 +2524,7 @@ public void testNameNodeCreateSnapshotTrashRootOnStartup()
MiniDFSCluster cluster =
new MiniDFSCluster.Builder(conf).numDataNodes(1).build();
try {
final DistributedFileSystem dfs = cluster.getFileSystem();
DistributedFileSystem dfs = cluster.getFileSystem();
final Path testDir = new Path("/disallowss/test2/");
final Path file0path = new Path(testDir, "file-0");
dfs.create(file0path).close();
@ -2535,7 +2535,20 @@ public void testNameNodeCreateSnapshotTrashRootOnStartup()
// Set dfs.namenode.snapshot.trashroot.enabled=true
conf.setBoolean("dfs.namenode.snapshot.trashroot.enabled", true);
cluster.setNameNodeConf(0, conf);
cluster.shutdown();
conf.setInt(DFSConfigKeys.DFS_NAMENODE_SAFEMODE_EXTENSION_KEY, 0);
conf.setInt(DFSConfigKeys.DFS_NAMENODE_SAFEMODE_MIN_DATANODES_KEY, 1);
cluster.restartNameNode(0);
dfs = cluster.getFileSystem();
assertTrue(cluster.getNameNode().isInSafeMode());
// Check .Trash existence, won't be created now
assertFalse(dfs.exists(trashRoot));
// Start a datanode
cluster.startDataNodes(conf, 1, true, null, null);
// Wait long enough for safemode check to retire
try {
Thread.sleep(1000);
} catch (InterruptedException ignored) {}
// Check .Trash existence, should be created now
assertTrue(dfs.exists(trashRoot));
// Check permission
@ -2553,4 +2566,6 @@ public void testNameNodeCreateSnapshotTrashRootOnStartup()
}
}
}
}