You are viewing a plain text version of this content. The canonical link for it is here.

Posted to issues@flink.apache.org by "xinyang gao (JIRA)" <ji...@apache.org> on 2017/02/20 11:24:44 UTC

[jira] [Updated] (FLINK-5847) Not able to deploy Flink 1.3-SNAPSHOT on Openshift (Platform as a Service)

     [ https://issues.apache.org/jira/browse/FLINK-5847?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

xinyang gao updated FLINK-5847:
-------------------------------
    Summary: Not able to deploy Flink 1.3-SNAPSHOT on Openshift (Platform as a Service)  (was: Not able to deploy Flink 1.3-SNAPSHOT on Openshift Pass)

> Not able to deploy Flink 1.3-SNAPSHOT on Openshift (Platform as a Service)
> --------------------------------------------------------------------------
>
>                 Key: FLINK-5847
>                 URL: https://issues.apache.org/jira/browse/FLINK-5847
>             Project: Flink
>          Issue Type: Bug
>          Components: JobManager, Local Runtime
>    Affects Versions: 1.3.0
>         Environment: Openshift Paas
>            Reporter: xinyang gao
>            Priority: Blocker
>              Labels: easyfix
>             Fix For: 1.3.0
>
>
> I am now trying to deploy Flink-1.3-Snapshot on Openshift which is a docker-environment. I was able to successfully deploy Flink-1.1.2 version on Openshift although I encountered the exception that "org.apache.flink.runtime.util.EnvironmentInformation          - Error while accessing user/group information via Hadoop utils.java.io.IOException: failure to login", this exception was handled and finally the jobmanager can be fired up on Openshift.
> However when it comes 1.3-SNAPSHOT, it will finally incur the error that "Exception in thread "main" java.lang.Exception: unable to establish the security context" which is caused by "Caused by: javax.security.auth.login.LoginException: java.lang.NullPointerException: invalid null input: name", then the Jobmanager failed to start. The reason looks like is that containers run on openshift are assigned a random uid rather than running as the default uid declared by the image. this can cause problems for some images. So for the container it does not have a current user name, however due to some reason in Flink 1.3-SNAPSHOT Hadoop needs to extract the user name from UNIX, if null it will throw an exception and stop firing up.



--
This message was sent by Atlassian JIRA
(v6.3.15#6346)