[ https://issues.apache.org/jira/browse/CASSANDRA-13368?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15938645#comment-15938645 ]
William R. Speirs commented on CASSANDRA-13368: ----------------------------------------------- [~spo...@gmail.com] hu, that's pretty curious then. I double-checked and Cassandra is using SLF4J v1.7.2. From my logs for example: {noformat} ERROR [STREAM-OUT-/X.X.X.X] 2017-03-19 20:29:15,224 StreamSession.java:512 - [Stream #ac1424f0-0cfd-11e7-b5a3-cb3016d1596d] Streaming error occurred {noformat} That code is: {noformat} 501 public void onError(Throwable e) 502 { 503 if (e instanceof SocketTimeoutException) 504 { 505 logger.error("[Stream #{}] Streaming socket timed out. This means the session peer stopped responding or " + 506 "is still processing received data. If there is no sign of failure in the other end or a very " + 507 "dense table is being transferred you may want to increase streaming_socket_timeout_in_ms " + 508 "property. Current value is {}ms.", planId(), DatabaseDescriptor.getStreamingSocketTimeout(), e); 509 } 510 else 511 { 512 logger.error("[Stream #{}] Streaming error occurred", planId(), e); 513 } 514 // send session failure message 515 if (handler.isOutgoingConnected()) 516 handler.sendMessage(new SessionFailedMessage()); 517 // fail session 518 closeSession(State.FAILED); 519 } {noformat} So I'm at a loss as to why {{e}} is not being properly interpreted as {{Throwable}} and therefore printing the stack trace. Thoughts? > Exception Stack not Printed as Intended in Error Logs > ----------------------------------------------------- > > Key: CASSANDRA-13368 > URL: https://issues.apache.org/jira/browse/CASSANDRA-13368 > Project: Cassandra > Issue Type: Bug > Reporter: William R. Speirs > Priority: Trivial > Labels: lhf > Fix For: 2.1.x > > Attachments: cassandra-13368-2.1.patch > > > There are a number of instances where it appears the programmer intended to > print a stack trace in an error message, but it is not actually being > printed. For example, in {{BlacklistedDirectories.java:54}}: > {noformat} > catch (Exception e) > { > JVMStabilityInspector.inspectThrowable(e); > logger.error("error registering MBean {}", MBEAN_NAME, e); > //Allow the server to start even if the bean can't be registered > } > {noformat} > The logger will use the second argument for the braces, but will ignore the > exception {{e}}. It would be helpful to have the stack traces of these > exceptions printed. I propose adding a second line that prints the full stack > trace: {{logger.error(e.getMessage(), e);}} > On the 2.1 branch, I found 8 instances of these types of messages: > {noformat} > db/BlacklistedDirectories.java:54: logger.error("error registering > MBean {}", MBEAN_NAME, e); > io/sstable/SSTableReader.java:512: logger.error("Corrupt sstable > {}; skipped", descriptor, e); > net/OutboundTcpConnection.java:228: logger.error("error > processing a message intended for {}", poolReference.endPoint(), e); > net/OutboundTcpConnection.java:314: logger.error("error > writing to {}", poolReference.endPoint(), e); > service/CassandraDaemon.java:231: logger.error("Exception in > thread {}", t, e); > service/CassandraDaemon.java:562: logger.error("error > registering MBean {}", MBEAN_NAME, e); > streaming/StreamSession.java:512: logger.error("[Stream #{}] > Streaming error occurred", planId(), e); > transport/Server.java:442: logger.error("Problem retrieving > RPC address for {}", endpoint, e); > {noformat} > And one where it'll print the {{toString()}} version of the exception: > {noformat} > db/Directories.java:689: logger.error("Could not calculate the > size of {}. {}", input, e); > {noformat} > I'm happy to create a patch for each branch, just need a little guidance on > how to do so. We're currently running 2.1 so I started there. -- This message was sent by Atlassian JIRA (v6.3.15#6346)