HDFS-16870 HDFS client ip should also be logged when NameNode is processing reportBadBlocks #5237
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
slfan1989
reviewed
Dec 17, 2022
| for (int j = 0; j < nodes.length; j++) { | ||
| NameNode.stateChangeLog.info("*DIR* reportBadBlocks for block: {} on" | ||
| + " datanode: {}", blk, nodes[j].getXferAddr()); | ||
| + " datanode: {}" + " client: {}", blk, nodes[j].getXferAddr(), Server.getRemoteIp()); |
Contributor
There was a problem hiding this comment.
From my personal point of view, reportBadBlocks should be reported by DN. What does this have to do with the client? Server.getRemoteIp() is somewhat expensive, is there a good enough reason for us to do this?
Member
There was a problem hiding this comment.
Instead of Server.getRemoteIp(), do a getClientMachine() instead, it handles calls via RBF.
Fetching of this IP Address should be outside lock, doing this inside lock will hit performance.
You are logging this inside the loop, so better extract a variable outside.
reportBadBlocks is there in both ClientProtocol as well as in DatanodeProtocol. Can check somewhat here for pointers
Contributor
Author
There was a problem hiding this comment.
Instead of
Server.getRemoteIp(), do agetClientMachine()instead, it handles calls via RBF. Fetching of this IP Address should be outside lock, doing this inside lock will hit performance. You are logging this inside the loop, so better extract a variable outside.reportBadBlocks is there in both ClientProtocol as well as in DatanodeProtocol. Can check somewhat here for pointers
@ayushtkn @slfan1989 Thanks for review, I moved getClientIp out of the for loop and write lock.
and getClientMachine is not accessible here, this function always invoked by an open file.
|
💔 -1 overall
This message was automatically generated. |
…essing reportBadBlocks
Daniel-009497
force-pushed
the
HDFS-16870
branch
from
43d48f5 to
05137dd
Compare
|
💔 -1 overall
This message was automatically generated. |
Contributor
Author
|
@brahmareddybattula @ayushtkn |
ayushtkn
requested changes
Dec 31, 2022
Member
ayushtkn
left a comment
ayushtkn
left a comment
There was a problem hiding this comment.
One change post that changes LGTM
| */ | ||
| void reportBadBlocks(LocatedBlock[] blocks) throws IOException { | ||
| checkOperation(OperationCategory.WRITE); | ||
| InetAddress remoteIp = Server.getRemoteIp(); |
Member
There was a problem hiding this comment.
This ain't the correct way to do as I previously mentioned, use getClientMachine()
Something like this on top of your present changes
diff --git a/hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/server/namenode/FSNamesystem.java b/hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/server/namenode/FSNamesystem.java
index b624ab76cc0..ed95c912171 100644
--- a/hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/server/namenode/FSNamesystem.java
+++ b/hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/server/namenode/FSNamesystem.java
@@ -5882,9 +5882,8 @@ private INodeFile checkUCBlock(ExtendedBlock block,
/**
* Client is reporting some bad block locations.
*/
- void reportBadBlocks(LocatedBlock[] blocks) throws IOException {
+ void reportBadBlocks(String clientMachine, LocatedBlock[] blocks) throws IOException {
checkOperation(OperationCategory.WRITE);
- InetAddress remoteIp = Server.getRemoteIp();
writeLock();
try {
checkOperation(OperationCategory.WRITE);
@@ -5894,7 +5893,7 @@ void reportBadBlocks(LocatedBlock[] blocks) throws IOException {
String[] storageIDs = blocks[i].getStorageIDs();
for (int j = 0; j < nodes.length; j++) {
NameNode.stateChangeLog.info("*DIR* reportBadBlocks for block: {} on"
- + " datanode: {}" + " client: {}", blk, nodes[j].getXferAddr(), remoteIp);
+ + " datanode: {}" + " client: {}", blk, nodes[j].getXferAddr(), clientMachine);
blockManager.findAndMarkBlockAsCorrupt(blk, nodes[j],
storageIDs == null ? null: storageIDs[j],
"client machine reported it");
diff --git a/hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/server/namenode/NameNodeRpcServer.java b/hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/server/namenode/NameNodeRpcServer.java
index b19bfc13acf..eae945c7458 100644
--- a/hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/server/namenode/NameNodeRpcServer.java
+++ b/hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/server/namenode/NameNodeRpcServer.java
@@ -991,7 +991,7 @@ public boolean complete(String src, String clientName,
@Override // ClientProtocol, DatanodeProtocol
public void reportBadBlocks(LocatedBlock[] blocks) throws IOException {
checkNNStartup();
- namesystem.reportBadBlocks(blocks);
+ namesystem.reportBadBlocks(getClientMachine(), blocks);
}
@Override // ClientProtocol
Contributor
|
We're closing this stale PR because it has been open for 100 days with no activity. This isn't a judgement on the merit of the PR in any way. It's just a way of keeping the PR queue manageable. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
4 participants
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
There are two scenarios involded for reportBadBlocks
1-HDFS client will report bad block to NameNode once the block size or data is not consistent with meta;
2-DataNode will report bad block to NameNode via heartbeat if Replica stored on Datanode is corrupted or be modified.
As for now, when namenode process reportBadBlock rpc request, only DataNode address is logged.
Client Ip should also be logged to distinguish where the report comes from, which is very useful for trouble shooting.