Webhdfs output plugin does not create new files

Elastic Stack Logstash
#1

Hi,

we have problems writing to hdfs via the webhdfs plugin. It does not create new files but will wirte to a file that does already exist. This prevents using tags like date, time, host and so on.

Does someone have the same problem and how do I solve this?

Thanks,

Olaf.

#2

Providing your config as well as what version you are running would help.

#3

Hi,

sorry. the config is pretty vanilla. The version is logstash 2.1 with webhdfs pulled from github (was not included in the "all-plugins" version).

the error message was

webhdfs write caused an exception: {"RemoteException":{"message":"Append failed for file: 
\/services\/logstash\/test.log, error: No such file or directory
(2)","exception":"IOException","javaClassName":"java.io.IOException"}}. Maybe you should increase retry_interval      
or reduce number of workers. Retrying... {:level=>:warn, :file=>"logstash/outputs/webhdfs.rb", :line=>"191",   
:method=>"write_data"}

until I created the file with "hadoop dfs -touchz /services/logstash/test.log". Then it worked.

input {
 log4j {
  mode => server
  host => "0.0.0.0"
  port => "4560"
  type => "log4j"
  }
}

output {
 webhdfs {
  host => "355.305.404.230"         
  port => "14000"
  path => "/services/logstash/test.log"  
  user => "hadoop"
}
#4

Hi,

I too face similar issue. Using webhdfs plugin in logstash to push a copy to HDFS.
It does creates file with dynamic names[dd/mm/yy etc], but writes only first line. All subsequent logs that logstash tries to push results in same error as stated in this post.

When i use pre created file like test.log, it works smoothly. I went thru some posts and they are suggesting to close the HDFS file once created and reopen again for appending. I guess that is not working properly with the plugin.

Strangely when i make dfs.replication factor to 1, it does work in all scenarios. I am using HDP sandbox.

#5