Windows Hadoop client timeout exception connecting to Linux Hadoop cluster

[Jochen Hebbrecht <jo...@gmail.com>]

Hi,

This is the situation: a Hadoop cluster running on a Linux environment
(version 2.5.0 CDH 5.3.3). We can contact the cluster on a Linux client
with no issues and perform jobs on it.

{bash}
$ hadoop --config ../etc/myenvironment fs -ls /
{bash}
=> is working fine

On a Windows machine, we use exactly the same configuration and client, but
the connection to the server takes a very longs time and eventually gets
killed.

{bash}
$ hadoop --config ../etc/myenvironment/ fs -ls /
Java config name: null
Native config name: C:\WINDOWS\krb5.ini
2016-07-11 15:25:21,138 INFO  [main] ipc.Client
(Client.java:handleConnectionTimeout(814)) - Retrying connect to server:
XXX:8020. Already tried 0 time(s); maxRetries=45
2016-07-11 15:25:41,138 INFO  [main] ipc.Client
(Client.java:handleConnectionTimeout(814)) - Retrying connect to server:
XXX:8020. Already tried 1 time(s); maxRetries=45
2016-07-11 15:26:01,156 INFO  [main] ipc.Client
(Client.java:handleConnectionTimeout(814)) - Retrying connect to server:
XXX:8020. Already tried 2 time(s); maxRetries=45
...
{bash}

I think I'm running into this issue:
https://discuss.zendesk.com/hc/en-us/articles/203758976-hdfs-dfs-ls-commands-will-hang-indefinitely-when-secure-hdfs-is-enabled
Although my principal looks OK:

{bash}
$ klist
Ticket cache: FILE:/tmp/krb5cc_1291429
Default principal: myusername@mydomain

Valid starting       Expires              Service principal
07/11/2016 15:03:00  07/12/2016 01:03:00  krbtgt/mydomain@mydomainname
        renew until 07/11/2016 15:03:00
{bash}

Any idea how I can debug this?

Kind regards,
Jochen Hebbrecht