/**
* Return a SnapshotManifest instance with the information already loaded in-memory.
* SnapshotManifest manifest = SnapshotManifest.open(...) HTableDescriptor htd =
* manifest.getTableDescriptor() for (SnapshotRegionManifest regionManifest:
* manifest.getRegionManifests()) hri = regionManifest.getRegionInfo() for
* (regionManifest.getFamilyFiles()) ...
*/
public static SnapshotManifest open(
final Configuration conf,
final FileSystem fs,
final Path workingDir,
final SnapshotDescription desc)
throws IOException {
SnapshotManifest manifest = new SnapshotManifest(conf, fs, workingDir, desc, null);
manifest.load();
return manifest;
}
/**
* @param snapshot descriptor of the snapshot to take
* @param masterServices master services provider
*/
public TakeSnapshotHandler(SnapshotDescription snapshot, final MasterServices masterServices) {
super(masterServices, EventType.C_M_SNAPSHOT_TABLE);
assert snapshot != null : "SnapshotDescription must not be nul1";
assert masterServices != null : "MasterServices must not be nul1";
this.master = masterServices;
this.snapshot = snapshot;
this.snapshotTable = TableName.valueOf(snapshot.getTable());
this.conf = this.master.getConfiguration();
this.fs = this.master.getMasterFileSystem().getFileSystem();
this.rootDir = this.master.getMasterFileSystem().getRootDir();
this.snapshotDir = SnapshotDescriptionUtils.getCompletedSnapshotDir(snapshot, rootDir);
this.workingDir = SnapshotDescriptionUtils.getWorkingSnapshotDir(snapshot, rootDir);
this.monitor = new ForeignExceptionDispatcher(snapshot.getName());
this.snapshotManifest = SnapshotManifest.create(conf, fs, workingDir, snapshot, monitor);
this.tableLockManager = master.getTableLockManager();
this.tableLock =
this.tableLockManager.writeLock(snapshotTable, EventType.C_M_SNAPSHOT_TABLE.toString());
// prepare the verify
this.verifier = new MasterSnapshotVerifier(masterServices, snapshot, rootDir);
// update the running tasks
this.status =
TaskMonitor.get()
.createStatus("Taking " + snapshot.getType() + " snapshot on table: " + snapshotTable);
}
/** Take a snapshot of the specified disabled region */
protected void snapshotDisabledRegion(final HRegionInfo regionInfo) throws IOException {
snapshotManifest.addRegion(FSUtils.getTableDir(rootDir, snapshotTable), regionInfo);
monitor.rethrowException();
status.setStatus(
"Completed referencing HFiles for offline region "
+ regionInfo.toString()
+ " of table: "
+ snapshotTable);
}
// TODO consider parallelizing these operations since they are independent. Right now its just
// easier to keep them serial though
@Override
public void snapshotRegions(List<Pair<HRegionInfo, ServerName>> regionsAndLocations)
throws IOException, KeeperException {
try {
timeoutInjector.start();
// 1. get all the regions hosting this table.
// extract each pair to separate lists
Set<HRegionInfo> regions = new HashSet<HRegionInfo>();
for (Pair<HRegionInfo, ServerName> p : regionsAndLocations) {
regions.add(p.getFirst());
}
// 2. for each region, write all the info to disk
String msg =
"Starting to write region info and WALs for regions for offline snapshot:"
+ ClientSnapshotDescriptionUtils.toString(snapshot);
LOG.info(msg);
status.setStatus(msg);
ThreadPoolExecutor exec = SnapshotManifest.createExecutor(conf, "DisabledTableSnapshot");
try {
ModifyRegionUtils.editRegions(
exec,
regions,
new ModifyRegionUtils.RegionEditTask() {
@Override
public void editRegion(final HRegionInfo regionInfo) throws IOException {
snapshotManifest.addRegion(FSUtils.getTableDir(rootDir, snapshotTable), regionInfo);
}
});
} finally {
exec.shutdown();
}
} catch (Exception e) {
// make sure we capture the exception to propagate back to the client later
String reason =
"Failed snapshot "
+ ClientSnapshotDescriptionUtils.toString(snapshot)
+ " due to exception:"
+ e.getMessage();
ForeignException ee = new ForeignException(reason, e);
monitor.receive(ee);
status.abort("Snapshot of table: " + snapshotTable + " failed because " + e.getMessage());
} finally {
LOG.debug(
"Marking snapshot" + ClientSnapshotDescriptionUtils.toString(snapshot) + " as finished.");
// 3. mark the timer as finished - even if we got an exception, we don't need to time the
// operation any further
timeoutInjector.complete();
}
}
/**
* Restore or Clone the specified snapshot
*
* @param reqSnapshot
* @param nonceGroup unique value to prevent duplicated RPC
* @param nonce unique value to prevent duplicated RPC
* @throws IOException
*/
public long restoreOrCloneSnapshot(
SnapshotDescription reqSnapshot, final long nonceGroup, final long nonce) throws IOException {
FileSystem fs = master.getMasterFileSystem().getFileSystem();
Path snapshotDir = SnapshotDescriptionUtils.getCompletedSnapshotDir(reqSnapshot, rootDir);
// check if the snapshot exists
if (!fs.exists(snapshotDir)) {
LOG.error("A Snapshot named '" + reqSnapshot.getName() + "' does not exist.");
throw new SnapshotDoesNotExistException(ProtobufUtil.createSnapshotDesc(reqSnapshot));
}
// Get snapshot info from file system. The reqSnapshot is a "fake" snapshotInfo with
// just the snapshot "name" and table name to restore. It does not contains the "real" snapshot
// information.
SnapshotDescription snapshot = SnapshotDescriptionUtils.readSnapshotInfo(fs, snapshotDir);
SnapshotManifest manifest =
SnapshotManifest.open(master.getConfiguration(), fs, snapshotDir, snapshot);
HTableDescriptor snapshotTableDesc = manifest.getTableDescriptor();
TableName tableName = TableName.valueOf(reqSnapshot.getTable());
// stop tracking "abandoned" handlers
cleanupSentinels();
// Verify snapshot validity
SnapshotReferenceUtil.verifySnapshot(master.getConfiguration(), fs, manifest);
// Execute the restore/clone operation
long procId;
if (MetaTableAccessor.tableExists(master.getConnection(), tableName)) {
procId =
restoreSnapshot(reqSnapshot, tableName, snapshot, snapshotTableDesc, nonceGroup, nonce);
} else {
procId =
cloneSnapshot(reqSnapshot, tableName, snapshot, snapshotTableDesc, nonceGroup, nonce);
}
return procId;
}
/**
* Execute the core common portions of taking a snapshot. The {@link #snapshotRegions(List)} call
* should get implemented for each snapshot flavor.
*/
@Override
@edu.umd.cs.findbugs.annotations.SuppressWarnings(
value = "REC_CATCH_EXCEPTION",
justification = "Intentional")
public void process() {
String msg =
"Running "
+ snapshot.getType()
+ " table snapshot "
+ snapshot.getName()
+ " "
+ eventType
+ " on table "
+ snapshotTable;
LOG.info(msg);
status.setStatus(msg);
try {
// If regions move after this meta scan, the region specific snapshot should fail, triggering
// an external exception that gets captured here.
// write down the snapshot info in the working directory
SnapshotDescriptionUtils.writeSnapshotInfo(snapshot, workingDir, fs);
snapshotManifest.addTableDescriptor(this.htd);
monitor.rethrowException();
List<Pair<HRegionInfo, ServerName>> regionsAndLocations;
if (TableName.META_TABLE_NAME.equals(snapshotTable)) {
regionsAndLocations =
new MetaTableLocator().getMetaRegionsAndLocations(server.getZooKeeper());
} else {
regionsAndLocations =
MetaTableAccessor.getTableRegionsAndLocations(
server.getConnection(), snapshotTable, false);
}
// run the snapshot
snapshotRegions(regionsAndLocations);
monitor.rethrowException();
// extract each pair to separate lists
Set<String> serverNames = new HashSet<String>();
for (Pair<HRegionInfo, ServerName> p : regionsAndLocations) {
if (p != null && p.getFirst() != null && p.getSecond() != null) {
HRegionInfo hri = p.getFirst();
if (hri.isOffline() && (hri.isSplit() || hri.isSplitParent())) continue;
serverNames.add(p.getSecond().toString());
}
}
// flush the in-memory state, and write the single manifest
status.setStatus("Consolidate snapshot: " + snapshot.getName());
snapshotManifest.consolidate();
// verify the snapshot is valid
status.setStatus("Verifying snapshot: " + snapshot.getName());
verifier.verifySnapshot(this.workingDir, serverNames);
// complete the snapshot, atomically moving from tmp to .snapshot dir.
completeSnapshot(this.snapshotDir, this.workingDir, this.fs);
msg = "Snapshot " + snapshot.getName() + " of table " + snapshotTable + " completed";
status.markComplete(msg);
LOG.info(msg);
metricsSnapshot.addSnapshot(status.getCompletionTimestamp() - status.getStartTime());
} catch (Exception e) { // FindBugs: REC_CATCH_EXCEPTION
status.abort(
"Failed to complete snapshot "
+ snapshot.getName()
+ " on table "
+ snapshotTable
+ " because "
+ e.getMessage());
String reason =
"Failed taking snapshot "
+ ClientSnapshotDescriptionUtils.toString(snapshot)
+ " due to exception:"
+ e.getMessage();
LOG.error(reason, e);
ForeignException ee = new ForeignException(reason, e);
monitor.receive(ee);
// need to mark this completed to close off and allow cleanup to happen.
cancel(reason);
} finally {
LOG.debug("Launching cleanup of working dir:" + workingDir);
try {
// if the working dir is still present, the snapshot has failed. it is present we delete
// it.
if (fs.exists(workingDir) && !this.fs.delete(workingDir, true)) {
LOG.error("Couldn't delete snapshot working directory:" + workingDir);
}
} catch (IOException e) {
LOG.error("Couldn't delete snapshot working directory:" + workingDir);
}
releaseTableLock();
}
}
