[
https://issues.apache.org/jira/browse/HDFS-10986?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15566678#comment-15566678
]
Mingliang Liu commented on HDFS-10986:
--------------------------------------
Thanks for the review, [~xiaobingo].
{quote}
You might want to log it as warning or error msgs.
{code}234 LOG.info("stderr: " + err);{code}
{quote}
This is for test assistance readability. The stderr is expected instead of a
real test error. In v3 patch, I merged the stdout/stderr with the assertion
message; if the test fails, the failing message will have the detailed
stderr/stdout message:
{code}
assertTrue("Unexpected " + command + " stdout: " + out, outs.isEmpty());
assertTrue("Unexpected " + command + " stderr: " + err,
err.toString().contains("Exception"));
{code}
{quote}
java.lang.Exception: test timed out after 60000 milliseconds
{quote}
The reason was that the previous test in [HDFS-10972] is using 128.0.0.1 which
is an external IP address. We can not depend on such resource and assume the
1234 port there is not responding. In v3 patch, I keep the connection failure
case where we use the Xfer port instead of IPC port on localhost. The DFSAdmin
will get "Datanode unreachable. java.io.EOFException" immediately and the test
can pass locally.
> DFSAdmin should log detailed error message if any
> -------------------------------------------------
>
> Key: HDFS-10986
> URL: https://issues.apache.org/jira/browse/HDFS-10986
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: tools
> Reporter: Mingliang Liu
> Assignee: Mingliang Liu
> Attachments: HDFS-10986-branch-2.8.002.patch,
> HDFS-10986-branch-2.8.003.patch, HDFS-10986.000.patch, HDFS-10986.001.patch,
> HDFS-10986.002.patch, HDFS-10986.003.patch
>
>
> There are some subcommands in {{DFSAdmin}} that swallow IOException and give
> very limited error message, if any, to the stderr.
> {code}
> $ hdfs dfsadmin -getBalancerBandwidth 127.0.0.1:9866
> Datanode unreachable.
> $ hdfs dfsadmin -getDatanodeInfo localhost:9866
> Datanode unreachable.
> $ hdfs dfsadmin -evictWriters 127.0.0.1:9866
> $ echo $?
> -1
> {code}
> User is not able to get the exception stack even the LOG level is DEBUG. This
> is not very user friendly. Fortunately, if the port number is not accessible
> (say 9999), users can infer the detailed error message by IPC logs:
> {code}
> $ hdfs dfsadmin -getBalancerBandwidth 127.0.0.1:9999
> 2016-10-07 18:01:35,115 WARN util.NativeCodeLoader: Unable to load
> native-hadoop library for your platform... using builtin-java classes where
> applicable
> 2016-10-07 18:01:36,335 INFO ipc.Client: Retrying connect to server:
> localhost/127.0.0.1:9999. Already tried 0 time(s); retry policy is
> RetryUpToMaximumCountWithFixedSleep(maxRetries=10, sleepTime=1000
> MILLISECONDS)
> .....
> 2016-10-07 18:01:45,361 INFO ipc.Client: Retrying connect to server:
> localhost/127.0.0.1:9999. Already tried 9 time(s); retry policy is
> RetryUpToMaximumCountWithFixedSleep(maxRetries=10, sleepTime=1000
> MILLISECONDS)
> 2016-10-07 18:01:45,362 WARN ipc.Client: Failed to connect to server:
> localhost/127.0.0.1:9999: retries get failed due to exceeded maximum allowed
> retries number: 10
> java.net.ConnectException: Connection refused
> at sun.nio.ch.SocketChannelImpl.checkConnect(Native Method)
> at
> sun.nio.ch.SocketChannelImpl.finishConnect(SocketChannelImpl.java:717)
> ...
> at org.apache.hadoop.hdfs.tools.DFSAdmin.run(DFSAdmin.java:2073)
> at org.apache.hadoop.util.ToolRunner.run(ToolRunner.java:76)
> at org.apache.hadoop.util.ToolRunner.run(ToolRunner.java:90)
> at org.apache.hadoop.hdfs.tools.DFSAdmin.main(DFSAdmin.java:2225)
> Datanode unreachable.
> {code}
> We should fix this by providing detailed error message. Actually, the
> {{DFSAdmin#run}} already handles exception carefully, including:
> # set the exit ret value to -1
> # print the error message
> # log the exception stack trace (in DEBUG level)
> All we need to do is to not swallow exceptions without good reason.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: hdfs-issues-unsubscr...@hadoop.apache.org
For additional commands, e-mail: hdfs-issues-h...@hadoop.apache.org
