Re: failure in the STOP command

2016-12-02 Thread Gour Saha
Billie, this is a good catch. Ophir, I think you should make this small change and try your app stop again to see if it works. -Gour On 12/2/16, 10:13 AM, "Billie Rinaldi" wrote: >This subprocess.Popen does appear to be missing an env=env parameter: >https://github.com/apache/incubator-slider/

Re: failure in the STOP command

2016-12-02 Thread Ophir Etzion
I'm doing import params but it doesn't even reach there because it fails on `from resource_management import *` at the top of the file because the pythonpath isnt set. On Fri, Dec 2, 2016 at 1:09 PM, Gour Saha wrote: > Can you check your package script stop function if it is doing "import > para

Re: failure in the STOP command

2016-12-02 Thread Billie Rinaldi
This subprocess.Popen does appear to be missing an env=env parameter: https://github.com/apache/incubator-slider/blob/develop/slider-agent/src/main/python/agent/PythonExecutor.py#L153 On Fri, Dec 2, 2016 at 9:30 AM, Ophir Etzion wrote: > 1. you can't see the PYTHONPATH issue. you can see there i

Re: failure in the STOP command

2016-12-02 Thread Gour Saha
Can you check your package script stop function if it is doing "import params" like this - https://github.com/apache/incubator-slider/blob/develop/app-packages/hbase/ package/scripts/hbase_master.py#L48 If yes, then you might have to share your app-package scripts (without the app binary/tar), fo

Re: failure in the STOP command

2016-12-02 Thread Ophir Etzion
1. you can't see the PYTHONPATH issue. you can see there is no setting of the PYTHONPATH that you can see in the START command. 2. thanks for letting me know about release_timeout_secs but for my app I don't care if the containers die, the stop command sends an udp packet elsewhere. here is the ou

Re: failure in the STOP command

2016-12-02 Thread Gour Saha
Also keep in mind - if your application needs to run something useful when the stop cmd is initiated then you need to set an appropriate value to site.global.app_container.release_timeout_secs. Otherwise kill signals are sent to the agent containers via YARN (almost immediately) and the containers

Re: failure in the STOP command

2016-12-02 Thread Billie Rinaldi
It looks like the Traceback stack for the stop command output is truncated in the logs you pasted. I only see the first line of the Traceback: INFO 2016-11-30 18:07:03,919 PythonExecutor.py:97 - stop command output: err: Traceback (most recent call last): File "/export/hdk3/yarn/nm/usercache/hiv

Re: failure in the STOP command

2016-12-02 Thread Ophir Etzion
it does implement a STOP command that does something useful. it fails because the PYTHONPATH isn't set like it is in different commands. On Thu, Dec 1, 2016 at 10:38 PM, Gour Saha wrote: > Does enable_presto_worker_component.py support/implement a STOP command? > > Does your application need to

[jira] [Commented] (SLIDER-1174) Support Tensorflow on Slider

2016-12-02 Thread Yang Wang (JIRA)
[ https://issues.apache.org/jira/browse/SLIDER-1174?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15714803#comment-15714803 ] Yang Wang commented on SLIDER-1174: --- Cool, thanks a lot for your review. [~gsaha] > Su