Skip to content

[DRAFT]CommitActivityImpl: close FileSystem asynchronously to prevent DFSClient hang #4184

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
pratapaditya04 wants to merge 2 commits into
base: master
Choose a base branch
from
Open

[DRAFT]CommitActivityImpl: close FileSystem asynchronously to prevent DFSClient hang #4184

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
52cecc4
Use non-cached FileSystem in CommitActivityImpl to prevent DFSClient …
pratapaditya04 Apr 1, 2026
dfdc06e
making it async the fs closure
pratapaditya04 Apr 1, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
28 changes: 27 additions & 1 deletion ...poral/src/main/java/org/apache/gobblin/temporal/ddm/activity/impl/CommitActivityImpl.java
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -102,7 +102,9 @@ public CommitStats commit(WUProcessingSpec workSpec) {
int numDeserializationThreads = DEFAULT_NUM_DESERIALIZATION_THREADS;
Optional<String> optJobName = Optional.empty();
AutomaticTroubleshooter troubleshooter = null;
try (FileSystem fs = Help.loadFileSystem(workSpec)) {
FileSystem fs = null;
try {
fs = Help.loadFileSystem(workSpec);
JobState jobState = Help.loadJobState(workSpec, fs);

int heartBeatInterval = JobStateUtils.getHeartBeatInterval(jobState);
Expand Down Expand Up @@ -172,13 +174,37 @@ public CommitStats commit(WUProcessingSpec workSpec) {
new IOException(e)
);
} finally {
closeFileSystemAsync(fs);
String errCorrelator = String.format("Commit [%s]", calcCommitId(workSpec));
EventSubmitter eventSubmitter = workSpec.getEventSubmitterContext().create();
Help.finalizeTroubleshooting(troubleshooter, eventSubmitter, log, errCorrelator);
ExecutorsUtils.shutdownExecutorService(heartBeatExecutor, com.google.common.base.Optional.of(log));
}
}

/**
* Closes a {@link FileSystem} on a daemon thread to avoid blocking the caller.
*
* Using a non-cached (force-loaded) FileSystem means each commit activity owns its own {@link org.apache.hadoop.hdfs.DFSClient}
* and therefore its own IPC {@code Client}. Closing that Client via {@code Client.stop()} waits for all open
* connections to drain, which can hang indefinitely when connection threads are blocked in native socket reads.
* By closing asynchronously we keep the activity thread unblocked while cleanup proceeds in the background.
*/
private void closeFileSystemAsync(FileSystem fs) {
if (fs == null) {
return;
}
Thread t = new Thread(() -> {
try {
fs.close();
} catch (IOException e) {
log.warn("Async FileSystem close encountered an error (fs: {})", fs.getUri(), e);
}
}, "CommitActivity-AsyncFSClose");
t.setDaemon(true);
t.start();
}

/**
* Commit task states to the dataset state store.
* @param jobState
Expand Down