You are viewing a plain text version of this content. The canonical link for it is here.
Posted to common-commits@hadoop.apache.org by we...@apache.org on 2020/03/14 02:02:03 UTC
[hadoop] branch trunk updated: HDFS-14820. The default 8KB buffer
of BlockReaderRemote#newBlockReader#BufferedOutputStream is too big.
Contributed by Lisheng Sun.
This is an automated email from the ASF dual-hosted git repository.
weichiu pushed a commit to branch trunk
in repository https://gitbox.apache.org/repos/asf/hadoop.git
The following commit(s) were added to refs/heads/trunk by this push:
new 7b637a3 HDFS-14820. The default 8KB buffer of BlockReaderRemote#newBlockReader#BufferedOutputStream is too big. Contributed by Lisheng Sun.
7b637a3 is described below
commit 7b637a3341f2024470929b653a58e9a88dd036c0
Author: Wei-Chiu Chuang <we...@apache.org>
AuthorDate: Fri Mar 13 19:01:23 2020 -0700
HDFS-14820. The default 8KB buffer of BlockReaderRemote#newBlockReader#BufferedOutputStream is too big. Contributed by Lisheng Sun.
---
.../java/org/apache/hadoop/hdfs/client/HdfsClientConfigKeys.java | 2 +-
hadoop-hdfs-project/hadoop-hdfs/src/main/resources/hdfs-default.xml | 6 ++----
2 files changed, 3 insertions(+), 5 deletions(-)
diff --git a/hadoop-hdfs-project/hadoop-hdfs-client/src/main/java/org/apache/hadoop/hdfs/client/HdfsClientConfigKeys.java b/hadoop-hdfs-project/hadoop-hdfs-client/src/main/java/org/apache/hadoop/hdfs/client/HdfsClientConfigKeys.java
index 228d7ac..407462c 100755
--- a/hadoop-hdfs-project/hadoop-hdfs-client/src/main/java/org/apache/hadoop/hdfs/client/HdfsClientConfigKeys.java
+++ b/hadoop-hdfs-project/hadoop-hdfs-client/src/main/java/org/apache/hadoop/hdfs/client/HdfsClientConfigKeys.java
@@ -150,7 +150,7 @@ public interface HdfsClientConfigKeys {
TimeUnit.DAYS.toMillis(10); // 10 days
String DFS_CLIENT_BLOCK_READER_REMOTE_BUFFER_SIZE_KEY =
"dfs.client.block.reader.remote.buffer.size";
- int DFS_CLIENT_BLOCK_READER_REMOTE_BUFFER_SIZE_DEFAULT = 8192;
+ int DFS_CLIENT_BLOCK_READER_REMOTE_BUFFER_SIZE_DEFAULT = 512;
String DFS_CLIENT_DEAD_NODE_DETECTION_ENABLED_KEY =
"dfs.client.deadnode.detection.enabled";
diff --git a/hadoop-hdfs-project/hadoop-hdfs/src/main/resources/hdfs-default.xml b/hadoop-hdfs-project/hadoop-hdfs/src/main/resources/hdfs-default.xml
index ede3dc0..9fcc047 100755
--- a/hadoop-hdfs-project/hadoop-hdfs/src/main/resources/hdfs-default.xml
+++ b/hadoop-hdfs-project/hadoop-hdfs/src/main/resources/hdfs-default.xml
@@ -4305,13 +4305,11 @@
<property>
<name>dfs.client.block.reader.remote.buffer.size</name>
- <value>8192</value>
+ <value>512</value>
<description>
- The output stream buffer size of a DFSClient remote read. The buffer default value is 8KB. The buffer includes
+ The output stream buffer size of a DFSClient remote read. The buffer default value is 512B. The buffer includes
only some request parameters that are: block, blockToken, clientName, startOffset, len, verifyChecksum,
cachingStrategy.
- It is recommended to adjust the value according to the workload, which can reduce unnecessary memory
- usage and the frequency of the garbage collection. A value of 512 might be reasonable.
</description>
</property>
---------------------------------------------------------------------
To unsubscribe, e-mail: common-commits-unsubscribe@hadoop.apache.org
For additional commands, e-mail: common-commits-help@hadoop.apache.org