Close connection manager on current thread in RemoteClusterConnection #44805
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
ywelsch
added
>test
Collaborator
|
Pinging @elastic/es-distributed |
server/src/test/java/org/elasticsearch/transport/RemoteClusterConnectionTests.java
Show resolved
Hide resolved
| Iterator<Map.Entry<DiscoveryNode, Transport.Connection>> iterator = connectedNodes.entrySet().iterator(); | ||
| while (iterator.hasNext()) { | ||
| Map.Entry<DiscoveryNode, Transport.Connection> next = iterator.next(); | ||
| if (waitForPendingConnections) { |
Contributor
There was a problem hiding this comment.
Not sure if this is actually safe. If the problem is that we're not safely handling all the pending connections here because of some race with the thread-pool then we're leaking FDs and now have no visibility on that?
Can't we just do:
diff --git a/server/src/main/java/org/elasticsearch/transport/ConnectionManager.java b/server/src/main/java/org/elasticsearch/transport/ConnectionManager.java
index ed26d0b07cd..d87aca31a13 100644
--- a/server/src/main/java/org/elasticsearch/transport/ConnectionManager.java
+++ b/server/src/main/java/org/elasticsearch/transport/ConnectionManager.java
@@ -41,6 +41,7 @@ import java.util.Set;
import java.util.concurrent.ConcurrentMap;
import java.util.concurrent.CopyOnWriteArrayList;
import java.util.concurrent.CountDownLatch;
+import java.util.concurrent.TimeUnit;
import java.util.concurrent.atomic.AtomicBoolean;
/**
@@ -255,8 +256,17 @@ public class ConnectionManager implements Closeable {
try {
closeLatch.await();
} catch (InterruptedException e) {
- Thread.currentThread().interrupt();
- throw new IllegalStateException(e);
+ logger.warn("Interrupted while waiting for in-progress connections to finish.");
+ try {
+ if (closeLatch.await(30L, TimeUnit.SECONDS) == false) {
+ Thread.currentThread().interrupt();
+ throw new IllegalStateException(e);
+ }
+ } catch (InterruptedException ie) {
+ Thread.currentThread().interrupt();
+ e.addSuppressed(ie);
+ throw new IllegalStateException(e);
+ }
}
Iterator<Map.Entry<DiscoveryNode, Transport.Connection>> iterator = connectedNodes.entrySet().iterator();
while (iterator.hasNext()) {something like this? If we in fact are safely shutting down those connections eventually even if the threadpool is shut down too early this should still finish properly. If not, we get proper visibility and know we have to fix the shutdown sequence to not leak connection attempts here?
Contributor
There was a problem hiding this comment.
Actually looking at the linked failures in the issues, we have interrupt exceptions on connection establishment as well. I'll have to look into this some more after the meeting now, but I don't think we should just suppress the issue here.
Contributor
There was a problem hiding this comment.
@yannick see this one:
1> [2019-07-15T04:03:40,834][INFO ][o.e.t.TransportService ] [testLazyResolveTransportAddress] publish_address {127.0.0.1:14002}, bound_addresses {127.0.0.1:14002}, {[::1]:14002}
1> [2019-07-15T04:03:40,848][WARN ][o.e.t.RemoteClusterConnection] [org.elasticsearch.transport.RemoteClusterConnectionTests] fetching nodes from external cluster [test-cluster] failed
1> org.elasticsearch.common.util.CancellableThreads$ExecutionCancelledException: operation was cancelled reason [connect handler is closed]
1> at org.elasticsearch.common.util.CancellableThreads.checkForCancel(CancellableThreads.java:65) ~[main/:8.0.0-SNAPSHOT]
1> at org.elasticsearch.common.util.CancellableThreads.executeIO(CancellableThreads.java:130) ~[main/:8.0.0-SNAPSHOT]
1> at org.elasticsearch.transport.RemoteClusterConnection$ConnectHandler.collectRemoteNodes(RemoteClusterConnection.java:443) [main/:?]
1> at org.elasticsearch.transport.RemoteClusterConnection$ConnectHandler$1.doRun(RemoteClusterConnection.java:431) [main/:?]
1> at org.elasticsearch.common.util.concurrent.AbstractRunnable.run(AbstractRunnable.java:37) [main/:?]
1> at java.util.concurrent.Executors$RunnableAdapter.call(Executors.java:515) [?:?]
1> at java.util.concurrent.FutureTask.run(FutureTask.java:264) [?:?]
1> at org.elasticsearch.common.util.concurrent.ThreadContext$ContextPreservingRunnable.run(ThreadContext.java:699) [main/:?]
1> at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1128) [?:?]
1> at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:628) [?:?]
1> at java.lang.Thread.run(Thread.java:834) [?:?]
1> Suppressed: org.elasticsearch.transport.ConnectTransportException: [seed_node][127.0.0.1:14000] general node connection failure
1> at org.elasticsearch.transport.TcpTransport$ChannelsConnectedListener.lambda$onResponse$2(TcpTransport.java:948) ~[main/:?]
1> at org.elasticsearch.action.ActionListener$1.onFailure(ActionListener.java:70) ~[main/:?]
1> at org.elasticsearch.transport.TransportHandshaker$HandshakeResponseHandler.handleLocalException(TransportHandshaker.java:155) ~[main/:?]
1> at org.elasticsearch.transport.TransportHandshaker.lambda$sendHandshake$0(TransportHandshaker.java:67) ~[main/:?]
1> at org.elasticsearch.action.ActionListener.lambda$wrap$0(ActionListener.java:131) ~[main/:?]
1> at org.elasticsearch.action.ActionListener$1.onResponse(ActionListener.java:62) ~[main/:?]
1> at org.elasticsearch.action.ActionListener.lambda$toBiConsumer$3(ActionListener.java:159) ~[main/:?]
1> at org.elasticsearch.common.concurrent.CompletableContext.lambda$addListener$0(CompletableContext.java:39) ~[elasticsearch-core-8.0.0-SNAPSHOT.jar:8.0.0-SNAPSHOT]
1> at java.util.concurrent.CompletableFuture.uniWhenComplete(CompletableFuture.java:859) ~[?:?]
1> at java.util.concurrent.CompletableFuture$UniWhenComplete.tryFire(CompletableFuture.java:837) ~[?:?]
1> at java.util.concurrent.CompletableFuture.postComplete(CompletableFuture.java:506) ~[?:?]
1> at java.util.concurrent.CompletableFuture.complete(CompletableFuture.java:2073) ~[?:?]
1> at org.elasticsearch.common.concurrent.CompletableContext.complete(CompletableContext.java:61) ~[elasticsearch-core-8.0.0-SNAPSHOT.jar:8.0.0-SNAPSHOT]
1> at org.elasticsearch.nio.ChannelContext.closeFromSelector(ChannelContext.java:77) ~[elasticsearch-nio-8.0.0-SNAPSHOT.jar:8.0.0-SNAPSHOT]
1> at org.elasticsearch.nio.SocketChannelContext.closeFromSelector(SocketChannelContext.java:170) ~[elasticsearch-nio-8.0.0-SNAPSHOT.jar:8.0.0-SNAPSHOT]
1> at org.elasticsearch.nio.EventHandler.handleClose(EventHandler.java:220) ~[elasticsearch-nio-8.0.0-SNAPSHOT.jar:8.0.0-SNAPSHOT]
1> at org.elasticsearch.transport.nio.TestEventHandler.handleClose(TestEventHandler.java:199) ~[framework-8.0.0-SNAPSHOT.jar:8.0.0-SNAPSHOT]
1> at org.elasticsearch.nio.NioSelector.closeChannel(NioSelector.java:458) ~[elasticsearch-nio-8.0.0-SNAPSHOT.jar:8.0.0-SNAPSHOT]
1> at org.elasticsearch.nio.NioSelector.queueChannelClose(NioSelector.java:311) ~[elasticsearch-nio-8.0.0-SNAPSHOT.jar:8.0.0-SNAPSHOT]
1> at org.elasticsearch.nio.BytesChannelContext.closeChannel(BytesChannelContext.java:67) ~[elasticsearch-nio-8.0.0-SNAPSHOT.jar:8.0.0-SNAPSHOT]
2> jul. 15, 2019 4:03:51 A. M. com.carrotsearch.randomizedtesting.RandomizedRunner$QueueUncaughtExceptionsHandler uncaughtException
1> at org.elasticsearch.transport.nio.MockNioTransport$MockSocketChannel.close(MockNioTransport.java:304) ~[framework-8.0.0-SNAPSHOT.jar:8.0.0-SNAPSHOT]
1> at org.elasticsearch.core.internal.io.IOUtils.close(IOUtils.java:104) ~[elasticsearch-core-8.0.0-SNAPSHOT.jar:8.0.0-SNAPSHOT]
1> at org.elasticsearch.core.internal.io.IOUtils.close(IOUtils.java:86) ~[elasticsearch-core-8.0.0-SNAPSHOT.jar:8.0.0-SNAPSHOT]
1> at org.elasticsearch.common.network.CloseableChannel.closeChannels(CloseableChannel.java:89) ~[main/:?]
1> at org.elasticsearch.common.network.CloseableChannel.closeChannel(CloseableChannel.java:78) ~[main/:?]
2> WARNING: Uncaught exception in thread: Thread[elasticsearch[org.elasticsearch.transport.RemoteClusterConnectionTests][generic][T#3],5,TGRP-RemoteClusterConnectionTests]
2> java.lang.IllegalStateException: java.lang.InterruptedException
2> at __randomizedtesting.SeedInfo.seed([D538288F42486FCF]:0)
1> at org.elasticsearch.common.network.CloseableChannel.closeChannel(CloseableChannel.java:68) ~[main/:?]
2> at org.elasticsearch.transport.ConnectionManager.close(ConnectionManager.java:259)
2> at org.elasticsearch.common.util.concurrent.ThreadContext$ContextPreservingRunnable.run(ThreadContext.java:699)
1> at org.elasticsearch.transport.TcpTransport.onException(TcpTransport.java:591) ~[main/:?]
1> at org.elasticsearch.transport.nio.MockNioTransport.exceptionCaught(MockNioTransport.java:180) ~[framework-8.0.0-SNAPSHOT.jar:8.0.0-SNAPSHOT]
1> at org.elasticsearch.transport.nio.MockNioTransport$MockTcpChannelFactory.lambda$createChannel$2(MockNioTransport.java:214) ~[framework-8.0.0-SNAPSHOT.jar:8.0.0-SNAPSHOT]
1> at org.elasticsearch.nio.ChannelContext.handleException(ChannelContext.java:99) ~[elasticsearch-nio-8.0.0-SNAPSHOT.jar:8.0.0-SNAPSHOT]
2> at java.base/java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1128)
1> at org.elasticsearch.nio.EventHandler.readException(EventHandler.java:129) ~[elasticsearch-nio-8.0.0-SNAPSHOT.jar:8.0.0-SNAPSHOT]
1> at org.elasticsearch.transport.nio.TestEventHandler.readException(TestEventHandler.java:139) ~[framework-8.0.0-SNAPSHOT.jar:8.0.0-SNAPSHOT]
1> at org.elasticsearch.nio.NioSelector.handleRead(NioSelector.java:411) ~[elasticsearch-nio-8.0.0-SNAPSHOT.jar:8.0.0-SNAPSHOT]
2> at java.base/java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:628)
1> at org.elasticsearch.nio.NioSelector.processKey(NioSelector.java:246) ~[elasticsearch-nio-8.0.0-SNAPSHOT.jar:8.0.0-SNAPSHOT]
1> at org.elasticsearch.nio.NioSelector.singleLoop(NioSelector.java:174) ~[elasticsearch-nio-8.0.0-SNAPSHOT.jar:8.0.0-SNAPSHOT]
1> at org.elasticsearch.nio.NioSelector.runLoop(NioSelector.java:131) ~[elasticsearch-nio-8.0.0-SNAPSHOT.jar:8.0.0-SNAPSHOT]
2> at java.base/java.lang.Thread.run(Thread.java:834)
1> at java.lang.Thread.run(Thread.java:834) [?:?]
2> Caused by: java.lang.InterruptedException
It looks like we're failing to properly handle the connect handler failure here maybe and then just never release the ref?
original-brownbear
approved these changes
Jul 24, 2019
Contributor
original-brownbear
left a comment
original-brownbear
left a comment
There was a problem hiding this comment.
After discussing with @ywelsch on a another channel LGTM waiting for full close on the generic pool wasn't safe before either if the pool was already shut down anyway so this doesn't make things worse and fixes tests :)
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.
We have seen test failures on Windows (#44339) in RemoteClusterConnectionTests with the following stack trace:
Even though I don't have Windows to reproduce the issue, I think it's not specific to Windows but rather a timing-related issue. The problem is that RemoteClusterConnection closes the connection manager asynchronously, which races with the threadpool being shutdown at the end of the test.
Closes #44339, #44610