HDDS-834. Datanode goes OOM based because of segment size. Contributed by Mukul Kumar Singh.

This commit is contained in:
Shashikant Banerjee 2018-11-14 15:53:22 +05:30
parent 3fade865ce
commit a948281706
3 changed files with 24 additions and 9 deletions

View File

@ -74,7 +74,7 @@ public final class ScmConfigKeys {
public static final String DFS_CONTAINER_RATIS_SEGMENT_SIZE_KEY =
"dfs.container.ratis.segment.size";
public static final int DFS_CONTAINER_RATIS_SEGMENT_SIZE_DEFAULT =
1 * 1024 * 1024 * 1024;
16 * 1024;
public static final String DFS_CONTAINER_RATIS_SEGMENT_PREALLOCATED_SIZE_KEY =
"dfs.container.ratis.segment.preallocated.size";
public static final int

View File

@ -175,10 +175,10 @@
</property>
<property>
<name>dfs.container.ratis.segment.size</name>
<value>1073741824</value>
<value>16384</value>
<tag>OZONE, RATIS, PERFORMANCE</tag>
<description>The size of the raft segment used by Apache Ratis on datanodes.
(1 GB by default)
(16 KB by default)
</description>
</property>
<property>

View File

@ -120,7 +120,8 @@ public class ContainerStateMachine extends BaseStateMachine {
createContainerFutureMap;
private ExecutorService[] executors;
private final int numExecutors;
private final Map<Long, Long> containerCommandCompletionMap;
private final Map<Long, Long> applyTransactionCompletionMap;
private long lastIndex;
/**
* CSM metrics.
*/
@ -138,7 +139,8 @@ public ContainerStateMachine(RaftGroupId gid, ContainerDispatcher dispatcher,
this.executors = executors.toArray(new ExecutorService[numExecutors]);
this.writeChunkFutureMap = new ConcurrentHashMap<>();
this.createContainerFutureMap = new ConcurrentHashMap<>();
containerCommandCompletionMap = new ConcurrentHashMap<>();
applyTransactionCompletionMap = new ConcurrentHashMap<>();
this.lastIndex = RaftServerConstants.INVALID_LOG_INDEX;
}
@Override
@ -162,10 +164,12 @@ public void initialize(
private long loadSnapshot(SingleFileSnapshotInfo snapshot) {
if (snapshot == null) {
TermIndex empty = TermIndex.newTermIndex(0, 0);
TermIndex empty = TermIndex.newTermIndex(0,
RaftServerConstants.INVALID_LOG_INDEX);
LOG.info("The snapshot info is null." +
"Setting the last applied index to:" + empty);
setLastAppliedTermIndex(empty);
lastIndex = RaftServerConstants.INVALID_LOG_INDEX;
return RaftServerConstants.INVALID_LOG_INDEX;
}
@ -174,6 +178,7 @@ private long loadSnapshot(SingleFileSnapshotInfo snapshot) {
snapshot.getFile().getPath().toFile());
LOG.info("Setting the last applied index to " + last);
setLastAppliedTermIndex(last);
lastIndex = last.getIndex();
return last.getIndex();
}
@ -471,7 +476,7 @@ private void updateLastApplied() {
Long appliedTerm = null;
long appliedIndex = -1;
for(long i = getLastAppliedTermIndex().getIndex() + 1;; i++) {
final Long removed = containerCommandCompletionMap.remove(i);
final Long removed = applyTransactionCompletionMap.remove(i);
if (removed == null) {
break;
}
@ -479,7 +484,7 @@ private void updateLastApplied() {
appliedIndex = i;
}
if (appliedTerm != null) {
updateLastAppliedTermIndex(appliedIndex, appliedTerm);
updateLastAppliedTermIndex(appliedTerm, appliedIndex);
}
}
@ -489,6 +494,15 @@ private void updateLastApplied() {
@Override
public CompletableFuture<Message> applyTransaction(TransactionContext trx) {
long index = trx.getLogEntry().getIndex();
// ApplyTransaction call can come with an entryIndex much greater than
// lastIndex updated because in between entries in the raft log can be
// appended because raft config persistence. Just add a dummy entry
// for those.
for (long i = lastIndex + 1; i < index; i++) {
LOG.info("Gap in indexes at:{} detected, adding dummy entries ", i);
applyTransactionCompletionMap.put(i, trx.getLogEntry().getTerm());
}
try {
metrics.incNumApplyTransactionsOps();
ContainerCommandRequestProto requestProto =
@ -553,9 +567,10 @@ public CompletableFuture<Message> applyTransaction(TransactionContext trx) {
});
}
lastIndex = index;
future.thenAccept(m -> {
final Long previous =
containerCommandCompletionMap
applyTransactionCompletionMap
.put(index, trx.getLogEntry().getTerm());
Preconditions.checkState(previous == null);
updateLastApplied();