I am accessing HDFS using thrift.
This is the expected(and right) content on HDFS.
[hadoop@hdp-namenode-01 ~]$ hadoop fs -ls / Found 3 items drwxr-xr-x - hadoop supergroup 0 2012-04-26 14:07 /home drwxr-xr-x - hadoop supergroup 0 2012-04-26 14:21 /tmp drwxr-xr-x - hadoop supergroup 0 2012-04-26 14:20 /user
And then I start an HDFSThriftServer
[hadoop@hdp-namenode-01 ~]$ jps 17290 JobTracker 16980 NameNode 27289 Jps 17190 SecondaryNameNode 17511 RunJar 25270 HadoopThriftServer
Try to access content through thrift in PHP.
$transport = new TSocket(HDFS_HOST, HDFS_PORT); $transport->setRecvTimeout(60000); $transport->setSendTimeout(60000); $protocol =new TBinaryProtocol($transport); $client = new ThriftHadoopFileSystemClient($protocol); logv("connect hdfs"); $transport->open(); logv("testing existent of `%s'", $remote_uri); $remote_path = new Pathname(array('pathname' => $remote_uri)); $remote_file = null; try { $remote_file = $client->listStatus($remote_path); } catch(Exception $e) { } if (!$remote_file) loge("could not open `%s'", $remote_uri);
While $remote_uri
is an absolute path. For $remote_uri === '/non/existent'
or '/user'
, etc., listStatus always fail. But if I change it to '/tmp'
, I find that it lists the content of '/tmp'
of the thrift server's local FS.
So the returned content is local FS on the thrift server instead of HDFS! What's wrong here?