HBASE-28836 Parallelize the file archival to improve the split times #6243
+46
−9
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
… object store (s3)
|
@apurtell FYI |
| @@ -439,7 +439,7 @@ private static List<File> resolveAndArchive(FileSystem fs, Path baseArchiveDir, | |||
|
|
|||
| List<File> failures = new ArrayList<>(); | |||
| String startTime = Long.toString(start); | |||
| for (File file : toArchive) { | |||
| toArchive.parallelStream().forEach(file -> { | |||
There was a problem hiding this comment.
Good to check the difference between the parallelStream() and stream().parallel(). Java doc say that for parallelStream "It is allowable for this method to return a sequential stream."
| @@ -439,7 +439,7 @@ private static List<File> resolveAndArchive(FileSystem fs, Path baseArchiveDir, | |||
|
|
|||
| List<File> failures = new ArrayList<>(); | |||
| String startTime = Long.toString(start); | |||
| for (File file : toArchive) { | |||
| toArchive.parallelStream().forEach(file -> { | |||
There was a problem hiding this comment.
Also if toArchive is small, there are chances that making it parallel can take more time than doing the same sequentially.
There was a problem hiding this comment.
This wouldn't be the case as it is an IO operation (and diff wouldn't be noticable). So parallelizing it would improve it over what we already have. In pricinple i agree about what you are saying
There was a problem hiding this comment.
I thought parallelStream (and its underlying ForkJoinPool) is not really for I/O blocking operations.
There was a problem hiding this comment.
Yes, that's another point to consider
|
🎊 +1 overall
This message was automatically generated. |
|
💔 -1 overall
This message was automatically generated. |
|
Unit test failures seems to be unrelated. |
apurtell
approved these changes
Sep 13, 2024
virajjasani
reviewed
Sep 13, 2024
d-c-manning
reviewed
Sep 13, 2024
| @@ -439,7 +439,7 @@ private static List<File> resolveAndArchive(FileSystem fs, Path baseArchiveDir, | |||
|
|
|||
| List<File> failures = new ArrayList<>(); | |||
| String startTime = Long.toString(start); | |||
| for (File file : toArchive) { | |||
| toArchive.parallelStream().forEach(file -> { | |||
There was a problem hiding this comment.
I thought parallelStream (and its underlying ForkJoinPool) is not really for I/O blocking operations.
| @@ -439,7 +439,7 @@ private static List<File> resolveAndArchive(FileSystem fs, Path baseArchiveDir, | |||
|
|
|||
| List<File> failures = new ArrayList<>(); | |||
| String startTime = Long.toString(start); | |||
| for (File file : toArchive) { | |||
| toArchive.parallelStream().forEach(file -> { | |||
There was a problem hiding this comment.
is the access to ArrayList failures thread safe?
d-c-manning
reviewed
Sep 13, 2024
|
Since both region close as well as CompactedHFilesDischarger chore performs archival, we can keep Jira/PR title generic i.e. "Parallelize the file archival for performance improvement" or "Parallelize the file archival" (by removing both "split times" and "object stores" from it because the change is generic and not specific to object stores as per the PR change, and Jira description is good to explain why we need this specifically for S3). |
mnpoonia
changed the title
…times in object store (s3)" This reverts commit be6e570.
Apache9
reviewed
Sep 16, 2024
| @@ -439,7 +439,7 @@ private static List<File> resolveAndArchive(FileSystem fs, Path baseArchiveDir, | |||
|
|
|||
| List<File> failures = new ArrayList<>(); | |||
| String startTime = Long.toString(start); | |||
| for (File file : toArchive) { | |||
| toArchive.parallelStream().forEach(file -> { | |||
There was a problem hiding this comment.
There are IO operations in the code block so usually it is not safe to use parallelStream...
AFAIK parallelStream will make use of a common thread pool(ForkJoinPool.commonPool IIRC) which could be used by lots of other operations in java, so it is easy to introduce strange dead locks.
If we want to do it in parallel, we'd better introduce our own thread pool.
|
💔 -1 overall
This message was automatically generated. |
|
🎊 +1 overall
This message was automatically generated. |
|
nit: Parallize -> Parallelize |
mnpoonia
changed the title
|
I have added the logic for executor service doing the cleanup. I am still looking at executor service creation and config part. Basically trying to figure out the right config. Currently i am thinking of using |
|
🎊 +1 overall
This message was automatically generated. |
|
🎊 +1 overall
This message was automatically generated. |
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.
No description provided.