flink-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Prabhu V <vpra...@gmail.com>
Subject kerberos yarn - failure in long running streaming application
Date Mon, 14 Aug 2017 06:15:34 GMT
Hi,

I am running Flink-1.3.2 on yarn (Cloudera 2.6.0-cdh5.7.6). The application
stream data from kafka, groups by key, creates a session window and writes
to HDFS using a rich window function in the "window.apply" method.

The rich window function creates the sequence file thus

SequenceFile.createWriter(
                conf,
                new Option[] {
                        Writer.file(new Path("flink-output/" + filePath)),
                        Writer.compression(CompressionType.BLOCK,
                                new DefaultCodec()),
                        Writer.keyClass(BytesWritable.class),
                        Writer.valueClass(BytesWritable.class) })

The "conf" is created in the "open" method thus

conf = HadoopFileSystem.getHadoopConfiguration();
        for (Map.Entry<String, String> entry :
parameters.toMap().entrySet()) {
            conf.set(entry.getKey(), entry.getValue());
        }

where parameters is the flink.configuration.Configuration object that is an
argument to the open method

The applications runs for about 10 hours before it fails with kerberos
error "Caused by: javax.security.sasl.SaslException: GSS initiate failed
[Caused by GSSException: No valid credentials provided (Mechanism level:
Failed to find any Kerberos tgt)]"

The flink-conf.yaml has the following properties set.
security.kerberos.login.keytab: <keytab location>
security.kerberos.login.principal:<principal>
security.kerberos.login.contexts: Client,KafkaClien

Any help would be appreciated.


Thanks,
Prabhu

Mime
View raw message