HDFS-16593. Correct the BlocksRemoved metric on DataNode side #4353
Conversation
This file contains 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
|
💔 -1 overall
This message was automatically generated. |
|
@Hexiaoqiao Could you help me review this patch? The failed UTs not caused by this modification, and has been solved in other jira. |
|
💔 -1 overall
This message was automatically generated. |
|
The failed UT |
ZanderXu
changed the title
|
💔 -1 overall
This message was automatically generated. |
|
💔 -1 overall
This message was automatically generated. |
Hexiaoqiao
reviewed
Sep 1, 2022
| @@ -2350,6 +2350,9 @@ private void invalidate(String bpid, Block[] invalidBlks, boolean async) | |||
| removing = volumeMap.remove(bpid, invalidBlks[i]); | |||
| addDeletingBlock(bpid, removing.getBlockId()); | |||
| LOG.debug("Block file {} is to be deleted", removing.getBlockURI()); | |||
| if (datanode.getMetrics() != null) { | |||
There was a problem hiding this comment.
Ah, I believe that this will make BlocksRemoved more precisions. But I do not think it will be exact, consider that removing == null which means that DataNode do not manage this pending delete replica, so we should not increase the metric. Maybe there are some other cases, not thought deeply. I just suggest we should consider if removing will be null.
Another point, line 2352 will throw NPE but not decide if removing will null or not. This is another issue, not related to this PR.
There was a problem hiding this comment.
@Hexiaoqiao Thanks for your review. I learned a lot and have updated this patch, please help me review it again.
There was a problem hiding this comment.
@ZanderXu Sorry to pollute this flow. After review this context carefully, I realize removing will never be null actually. It will continue when meets null, and will never go to this logic (ref to Line2317). Another one, it is unnecessary to decide if datanode.getMetrics() is null or not. The first version is without any issues.
I have made a common mistake "ask a programmer to review 10 lines of code he'll find 10 issues, ask him to do 500 lines and he'll say it looks good."(via twitter)
"Wish every PR could be merged without any issues reviewers feedback." ^_^
There was a problem hiding this comment.
@ZanderXu Copy, sir.
Another one, it is unnecessary to decide if datanode.getMetrics() is null or not
We need it, because hadoop.hdfs.server.datanode.fsdataset.impl.TestFsDatasetImpl try to invalidate some blocks with null metrics. Or maybe we can modify the UT.
"Wish every PR could be merged without any issues reviewers feedback."
+1, and I will do my best to achieve it.
There was a problem hiding this comment.
We need it, because hadoop.hdfs.server.datanode.fsdataset.impl.TestFsDatasetImpl try to invalidate some blocks with null metrics.
I found that there are many datanode.getMetrics() calls without any null check in FsDatasetImpl, this is the same one. Not sure where will throw NPE, if that we should file another PR to fix them? Thanks.
There was a problem hiding this comment.
@Hexiaoqiao I have updated this code and fixed the failed UT org.apache.hadoop.hdfs.server.datanode.fsdataset.impl.TestFsDatasetImpl, please help me review it again. Thanks
|
💔 -1 overall
This message was automatically generated. |
Hexiaoqiao
approved these changes
Sep 2, 2022
|
💔 -1 overall
This message was automatically generated. |
|
🎊 +1 overall
This message was automatically generated. |
|
Committed to trunk. Thanks @ZanderXu . |
HarshitGupta11
pushed a commit
to HarshitGupta11/hadoop
that referenced
this pull request
Nov 28, 2022
…#4353). Contributed by ZanderXu. Signed-off-by: He Xiaoqiao <hexiaoqiao@apache.org>
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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.
Correct the BlocksRemoved metric on DataNode side.
When I read the code about
processCommandFromActive, I guess that the BlocksRemoved metric is imprecise.So I think we should increase the metric after DN successfully deleted the block: