Caleb Rackliffe created CASSANDRA-17663:
-------------------------------------------
Summary: Ensure that NettyStreamingMessageSender is not vulnerable
to triggering disk failure policy when file streaming tasks are interrupted
Key: CASSANDRA-17663
URL: https://issues.apache.org/jira/browse/CASSANDRA-17663
Project: Cassandra
Issue Type: Bug
Reporter: Caleb Rackliffe
Assignee: Caleb Rackliffe
This is very similar to what we fixed in CASSANDRA-17466 , just in a different
part of the repair/streaming machine.
{{NettyStreamingMessageSender}} is responsible for dispatching
{{FileStreamTask}} on an executor it manages to stream files to its peers. When
the sender is closed for any reason, like perhaps a peer blowing up while
deserializing the stream, the executor it manages is shut down w/ interruption
(i.e. {{shutdownNow()}}). This is problematic if we happen to have not gotten
very far along in {{FileStreamTask#run()}}. If we're just about to call
{{getOrCreateChannel()}}, which reads from the {{peers_v2}} system table, the
{{ChannelProxy}} read will throw a {{ClosedByInterruptedException}} and the
proxy will be useless. The twist is that, since CASSANDRA-15666, this exception
has essentially been swallowed by {{FileStreamTask}}'s exception handling. So
we don't see a {{ClosedByInterruptedException}} in the logs, but we do have
things like this pop up when anything else tries to hit the peers table:
{noformat}
ERROR 2022-05-19T21:49:23,218 [AntiEntropyStage:1]
org.apache.cassandra.service.CassandraDaemon:601 - Exception in thread
Thread[AntiEntropyStage:1,5,main]
java.lang.RuntimeException: FSReadError in
/cassandra/d3/data/system/peers_v2-c4325fbb8e5e3bafbd070f9250ed818e/system-peers_v2-nb-101-big-Data.db
at org.apache.cassandra.net.InboundSink.accept(InboundSink.java:108)
at org.apache.cassandra.net.InboundSink.accept(InboundSink.java:45)
at
org.apache.cassandra.net.InboundMessageHandler$ProcessMessage.run(InboundMessageHandler.java:433)
at
java.util.concurrent.Executors$RunnableAdapter.call(Executors.java:515)
at java.util.concurrent.FutureTask.run(FutureTask.java:264)
at
java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1128)
at
java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:628)
at
io.netty.util.concurrent.FastThreadLocalRunnable.run(FastThreadLocalRunnable.java:30)
[netty-all-4.1.58.Final.jar:4.1.58.Final]
at java.lang.Thread.run(Thread.java:834)
Caused by: org.apache.cassandra.io.FSReadError:
java.nio.channels.ClosedChannelException
at
org.apache.cassandra.io.util.ChannelProxy.read(ChannelProxy.java:143)
at
org.apache.cassandra.io.util.CompressedChunkReader$Standard.readChunk(CompressedChunkReader.java:115)
at
org.apache.cassandra.io.util.BufferManagingRebufferer.rebuffer(BufferManagingRebufferer.java:79)
at
org.apache.cassandra.io.util.RandomAccessReader.reBufferAt(RandomAccessReader.java:68)
at
org.apache.cassandra.io.util.RandomAccessReader.seek(RandomAccessReader.java:210)
at
org.apache.cassandra.io.util.FileHandle.createReader(FileHandle.java:151)
{noformat}
...and...
{noformat}
ERROR 2022-05-19T22:06:20,175 [CompactionExecutor:12]
org.apache.cassandra.service.CassandraDaemon:601 - Exception in thread
Thread[CompactionExecutor:12,1,main]
org.apache.cassandra.io.FSReadError: java.nio.channels.ClosedChannelException
at
org.apache.cassandra.io.util.ChannelProxy.read(ChannelProxy.java:143)
at
org.apache.cassandra.io.util.CompressedChunkReader$Standard.readChunk(CompressedChunkReader.java:115)
at
org.apache.cassandra.io.util.BufferManagingRebufferer.rebuffer(BufferManagingRebufferer.java:79)
at
org.apache.cassandra.io.util.RandomAccessReader.reBufferAt(RandomAccessReader.java:68)
at
org.apache.cassandra.io.util.RandomAccessReader.seek(RandomAccessReader.java:210)
at
org.apache.cassandra.io.sstable.format.big.BigTableScanner.seekToCurrentRangeStart(BigTableScanner.java:196)
at
org.apache.cassandra.io.sstable.format.big.BigTableScanner.access$400(BigTableScanner.java:52)
at
org.apache.cassandra.io.sstable.format.big.BigTableScanner$KeyScanningIterator.computeNext(BigTableScanner.java:305)
at
org.apache.cassandra.io.sstable.format.big.BigTableScanner$KeyScanningIterator.computeNext(BigTableScanner.java:282)
at
org.apache.cassandra.utils.AbstractIterator.hasNext(AbstractIterator.java:46)
at
org.apache.cassandra.io.sstable.format.big.BigTableScanner.hasNext(BigTableScanner.java:261)
{noformat}
...which obviously get us into trouble w/ the disk failure policy.
The fix proposed here is just to get the peers table read out of the thread
that can be interrupted. Specifically, {{NettyStreamingMessageSender}}
materializes a {{connectTo}} address at stream task creation time. This seemed
a better option than making shutdown non-interrupting, since that would mean
changing how the actual file streaming responds to shutdown.
--
This message was sent by Atlassian Jira
(v8.20.7#820007)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
