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

HDDS-1766. ContainerStateMachine is unable to increment lastAppliedTermIndex. Contributed by Mukul Kumar Singh. (#1072)

This commit is contained in:
Mukul Kumar Singh 2019-07-14 10:53:51 +05:30 committed by GitHub
parent 4a70a0d816
commit 0976f6fc30
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
1 changed files with 20 additions and 19 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

39
hadoop-hdds/container-service/src/main/java/org/apache/hadoop/ozone/container/common/transport/server/ratis/ContainerStateMachine.java
View File

@ -197,17 +197,16 @@ public class ContainerStateMachine extends BaseStateMachine {
if (snapshot == null) { if (snapshot == null) {
TermIndex empty = TermIndex empty =
TermIndex.newTermIndex(0, RaftLog.INVALID_LOG_INDEX); TermIndex.newTermIndex(0, RaftLog.INVALID_LOG_INDEX);
LOG.info( LOG.info("{}: The snapshot info is null. Setting the last applied index" +
"The snapshot info is null." + "Setting the last applied index to:" "to:{}", gid, empty);
+ empty);
setLastAppliedTermIndex(empty); setLastAppliedTermIndex(empty);
return RaftLog.INVALID_LOG_INDEX; return empty.getIndex();
} }
final File snapshotFile = snapshot.getFile().getPath().toFile(); final File snapshotFile = snapshot.getFile().getPath().toFile();
final TermIndex last = final TermIndex last =
SimpleStateMachineStorage.getTermIndexFromSnapshotFile(snapshotFile); SimpleStateMachineStorage.getTermIndexFromSnapshotFile(snapshotFile);
LOG.info("Setting the last applied index to " + last); LOG.info("{}: Setting the last applied index to {}", gid, last);
setLastAppliedTermIndex(last); setLastAppliedTermIndex(last);
// initialize the dispatcher with snapshot so that it build the missing // initialize the dispatcher with snapshot so that it build the missing
@ -243,18 +242,20 @@ public class ContainerStateMachine extends BaseStateMachine {
@Override @Override
public long takeSnapshot() throws IOException { public long takeSnapshot() throws IOException {
TermIndex ti = getLastAppliedTermIndex(); TermIndex ti = getLastAppliedTermIndex();
LOG.info("Taking snapshot at termIndex:" + ti); long startTime = Time.monotonicNow();
if (ti != null && ti.getIndex() != RaftLog.INVALID_LOG_INDEX) { if (ti != null && ti.getIndex() != RaftLog.INVALID_LOG_INDEX) {
final File snapshotFile = final File snapshotFile =
storage.getSnapshotFile(ti.getTerm(), ti.getIndex()); storage.getSnapshotFile(ti.getTerm(), ti.getIndex());
LOG.info("Taking a snapshot to file {}", snapshotFile); LOG.info("{}: Taking a snapshot at:{} file {}", gid, ti, snapshotFile);
try (FileOutputStream fos = new FileOutputStream(snapshotFile)) { try (FileOutputStream fos = new FileOutputStream(snapshotFile)) {
persistContainerSet(fos); persistContainerSet(fos);
} catch (IOException ioe) { } catch (IOException ioe) {
LOG.warn("Failed to write snapshot file \"" + snapshotFile LOG.info("{}: Failed to write snapshot at:{} file {}", gid, ti,
+ "\", last applied index=" + ti); snapshotFile);
throw ioe; throw ioe;
} }
LOG.info("{}: Finished taking a snapshot at:{} file:{} time:{}",
gid, ti, snapshotFile, (Time.monotonicNow() - startTime));
return ti.getIndex(); return ti.getIndex();
} }
return -1; return -1;
@ -337,7 +338,7 @@ public class ContainerStateMachine extends BaseStateMachine {
private ContainerCommandResponseProto dispatchCommand( private ContainerCommandResponseProto dispatchCommand(
ContainerCommandRequestProto requestProto, DispatcherContext context) { ContainerCommandRequestProto requestProto, DispatcherContext context) {
LOG.trace("dispatch {} containerID={} pipelineID={} traceID={}", LOG.trace("{}: dispatch {} containerID={} pipelineID={} traceID={}", gid,
requestProto.getCmdType(), requestProto.getContainerID(), requestProto.getCmdType(), requestProto.getContainerID(),
requestProto.getPipelineID(), requestProto.getTraceID()); requestProto.getPipelineID(), requestProto.getTraceID());
if (isBlockTokenEnabled) { if (isBlockTokenEnabled) {
@ -355,7 +356,7 @@ public class ContainerStateMachine extends BaseStateMachine {
} }
ContainerCommandResponseProto response = ContainerCommandResponseProto response =
dispatcher.dispatch(requestProto, context); dispatcher.dispatch(requestProto, context);
LOG.trace("response {}", response); LOG.trace("{}: response {}", gid, response);
return response; return response;
} }
@ -395,18 +396,18 @@ public class ContainerStateMachine extends BaseStateMachine {
.supplyAsync(() -> runCommand(requestProto, context), chunkExecutor); .supplyAsync(() -> runCommand(requestProto, context), chunkExecutor);
writeChunkFutureMap.put(entryIndex, writeChunkFuture); writeChunkFutureMap.put(entryIndex, writeChunkFuture);
LOG.debug("writeChunk writeStateMachineData : blockId " + write.getBlockID() LOG.debug(gid + ": writeChunk writeStateMachineData : blockId " +
+ " logIndex " + entryIndex + " chunkName " + write.getChunkData() write.getBlockID() + " logIndex " + entryIndex + " chunkName "
.getChunkName()); + write.getChunkData().getChunkName());
// Remove the future once it finishes execution from the // Remove the future once it finishes execution from the
// writeChunkFutureMap. // writeChunkFutureMap.
writeChunkFuture.thenApply(r -> { writeChunkFuture.thenApply(r -> {
metrics.incNumBytesWrittenCount( metrics.incNumBytesWrittenCount(
requestProto.getWriteChunk().getChunkData().getLen()); requestProto.getWriteChunk().getChunkData().getLen());
writeChunkFutureMap.remove(entryIndex); writeChunkFutureMap.remove(entryIndex);
LOG.debug("writeChunk writeStateMachineData completed: blockId " + write LOG.debug(gid + ": writeChunk writeStateMachineData completed: blockId" +
.getBlockID() + " logIndex " + entryIndex + " chunkName " + write write.getBlockID() + " logIndex " + entryIndex + " chunkName "
.getChunkData().getChunkName()); + write.getChunkData().getChunkName());
return r; return r;
}); });
return writeChunkFuture; return writeChunkFuture;
@ -564,12 +565,12 @@ public class ContainerStateMachine extends BaseStateMachine {
} }
} catch (Exception e) { } catch (Exception e) {
metrics.incNumReadStateMachineFails(); metrics.incNumReadStateMachineFails();
LOG.error("unable to read stateMachineData:" + e); LOG.error("{} unable to read stateMachineData:", gid, e);
return completeExceptionally(e); return completeExceptionally(e);
} }
} }
private void updateLastApplied() { private synchronized void updateLastApplied() {
Long appliedTerm = null; Long appliedTerm = null;
long appliedIndex = -1; long appliedIndex = -1;
for(long i = getLastAppliedTermIndex().getIndex() + 1;; i++) { for(long i = getLastAppliedTermIndex().getIndex() + 1;; i++) {