HDFS-16853. IPC shutdown failures. #5366
Closed
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
…LeaseRecoveryAfterNameNodeRestart failed
Trying to address the problem 1. MUST NOT submit into blocking queue while closing 2. MUST NOT call queue.put() in synchronous block. This design doesn't quite stop (2), though it should detect and warn if the problem surfaces. "Possible overlap in queue shutdown and request" No new tests; Change-Id: I2c7d3752bc8c4ab852015f53a5ef768d737efb2f
|
@virajjasani you were near this code...what do you think? @ZanderXu's core patch does the cleanup, but there's still a small window of possible overlap which I can't see how to get rid of through synchronized() blocks. I've got detection, but maybe some semaphore or similar needs to get involved so as to actually block cleanup while other threads are submitting work. dangerous though |
|
@xkrogen thoughts? |
|
💔 -1 overall
This message was automatically generated. |
virajjasani
reviewed
Feb 8, 2023
There was a problem hiding this comment.
Some minor comments, otherwise changes look good.
Not relevant mostly for this change but I wonder why we do not use running in addCall:
diff --git a/hadoop-common-project/hadoop-common/src/main/java/org/apache/hadoop/ipc/Client.java b/hadoop-common-project/hadoop-common/src/main/java/org/apache/hadoop/ipc/Client.java
index c0f90d98bc6..c0bef116e73 100644
--- a/hadoop-common-project/hadoop-common/src/main/java/org/apache/hadoop/ipc/Client.java
+++ b/hadoop-common-project/hadoop-common/src/main/java/org/apache/hadoop/ipc/Client.java
@@ -474,7 +474,7 @@ private void touch() {
* @return true if the call was added.
*/
private synchronized boolean addCall(Call call) {
- if (shouldCloseConnection.get())
+ if (shouldCloseConnection.get() || !running.get())
return false;
calls.put(call.id, call);
notify();
| */ | ||
| private synchronized SynchronousQueue<Pair<Call, ResponseBuffer>> acquireActiveRequestQueue() { | ||
| if (shouldCloseConnection.get() || !running.get()) { | ||
| LOG.debug("IPC client is stopped"); |
There was a problem hiding this comment.
nit: LOG.debug("IPC client {} is stopped", this) would be great
There was a problem hiding this comment.
i was thinking that myself
| * @return the queue or null. | ||
| */ | ||
| private synchronized SynchronousQueue<Pair<Call, ResponseBuffer>> acquireActiveRequestQueue() { | ||
| if (shouldCloseConnection.get() || !running.get()) { |
There was a problem hiding this comment.
We do not need additional synchronization on putLock object while accessing running correct?
Edit: This would be too much for accessing atomic boolean.
| // or it has happened but the finally {} clause has not been invoked (good). | ||
| // without knowing which, print a warning message so at least logs on | ||
| // a deadlock are meaningful. | ||
| LOG.warn("Possible overlap in queue shutdown and request"); |
There was a problem hiding this comment.
We can also print queueReservations.get() with this log
There was a problem hiding this comment.
ok. with the design of owens patch you should just have 1 thread in put() and one worker, but good to show.
+explicitly raising an exception if a call is made while closed not sure about the exception raising but when in, ipc tests fail, including the one on socket retries. Test problem? real problem? This gets complex fast, to complex for a last-minute fix. Change-Id: I86e738d22a927e36ac49c6c5d3e0fdb185416259 proposing: revert the HADOOP-18324 patch from 3.3.5
|
latest revision has a connection exception raised when queuing on a closed connection, rather than a silent no-op. this highlights that tests expecting socket retries to work now fail as stuff is going through the closed connection. Either the production code is wrong, or the mockito-based test isn't doing the right thing. it's too dangerous to try and get a fix in here for an RC, so I am proposing rolling back the big change and targeting the 3.3.6 release. |
|
see #5369 |
IMO that is reasonable thing to do given we are still not aware whether the impact is only limited to UTs. |
|
💔 -1 overall
This message was automatically generated. |
|
owen actually fixed this properly |
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.
Description of PR
Extension of #5162
Trying to address the problem
This design doesn't quite stop (2), though it should
detect and warn if the problem surfaces.
"Possible overlap in queue shutdown and request"
No new tests;
How was this patch tested?
For code changes:
LICENSE,LICENSE-binary,NOTICE-binaryfiles?