You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@ozone.apache.org by GitBox <gi...@apache.org> on 2021/11/30 09:15:57 UTC

[GitHub] [ozone] sadanand48 commented on a change in pull request #2866: HDDS-5851. Define a PutBlock/maxBuffer fixed boundary for streaming writes.

sadanand48 commented on a change in pull request #2866:
URL: https://github.com/apache/ozone/pull/2866#discussion_r759073315



##########
File path: hadoop-hdds/client/src/main/java/org/apache/hadoop/hdds/scm/storage/BlockDataStreamOutput.java
##########
@@ -257,13 +257,29 @@ public void write(ByteBuffer b, int off, int len) throws IOException {
     if (len == 0) {
       return;
     }
+    int curLen = len;
+    // set limit on the number of bytes that a ByteBuffer(StreamBuffer) can hold
+    int maxBufferLen = config.getDataStreamMaxBufferSize();
+    while (curLen > 0) {
+      int writeLen = Math.min(curLen, maxBufferLen);
+      final StreamBuffer buf = new StreamBuffer(b, off, writeLen);
+      off += writeLen;
+      bufferList.add(buf);
+      writeChunkToContainer(buf.duplicate());
+      curLen -= writeLen;
+      writtenDataLength += writeLen;
+      doFlushIfNeeded();
+    }
+  }
 
-    final StreamBuffer buf = new StreamBuffer(b, off, len);
-    bufferList.add(buf);
-
-    writeChunkToContainer(buf.duplicate());
-
-    writtenDataLength += len;
+  private void doFlushIfNeeded() throws IOException {
+    Preconditions.checkArgument(config.getDataStreamBufferFlushSize() > config
+        .getDataStreamMaxBufferSize());
+    long boundary = config.getDataStreamBufferFlushSize() / config
+        .getDataStreamMaxBufferSize();
+    if (bufferList.size() % boundary == 0) {
+      executePutBlock(false, false);

Review comment:
       Thanks @captainzmc for the comment. The plan here is to execute a putBlock at 32MB and not block the client right away by waiting for the flush futures and calling watchForCommit.  We will introduce another boundary say at 64MB to  check whether atleast 32MB is flushed and committed to all datanodes i.e watchForCommit() and waitForFlushFutures()  will occur at 64mb. (I will add this is in a different patch after this PR)




-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: issues-unsubscribe@ozone.apache.org

For queries about this service, please contact Infrastructure at:
users@infra.apache.org



---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscribe@ozone.apache.org
For additional commands, e-mail: issues-help@ozone.apache.org