Problems on setting up 0.2.3

[Yan...@nokia.com]

Hi,

 

First, I followed http://discoproject.org/doc/start/install.html#setup trying to set up a three-node cluster. It did not work when I tried to add nodes in the web interface. Then I followed http://discoproject.org/doc/start/troubleshoot.html to add nodes like [["localhost", "1"]] manually to disco_4441.config. But I still experienced some problems in the first paragraph in step 3) when I tried to run the sample job. The trace is copied below:

 

disco@jackalope:~/disco-0.2.3/examples$ python count_words.py http://localhost:8989

Starting Disco job..

Go to http://localhost:8989 to see status of the job.

Traceback (most recent call last):

  File "count_words.py", line 22, in <module>

    reduce = fun_reduce).wait()

  File "/usr/local/lib/python2.6/dist-packages/disco-0.1-py2.6.egg/disco/core.py", line 225, in g

    return f(*tuple([self.name] + list(args)), **kw)

  File "/usr/local/lib/python2.6/dist-packages/disco-0.1-py2.6.egg/disco/core.py", line 154, in wait

    status, results = self.results(name, timeout = poll_interval * 1000)

  File "/usr/local/lib/python2.6/dist-packages/disco-0.1-py2.6.egg/disco/core.py", line 132, in results

    results      = json.loads(self.request("/disco/ctrl/get_results", data))

  File "/usr/local/lib/python2.6/dist-packages/disco-0.1-py2.6.egg/disco/core.py", line 54, in request

    raise DiscoError('Got %s, make sure disco master is running at %s' % (e, self.host))

disco.error.DiscoError: Got HTTP exception (http://localhost:8989/disco/ctrl/get_results): Downloading http://localhost:8989/disco/ctrl/get_results failed after 10 attempts: HTTP exception (http://localhost:8989/disco/ctrl/get_results): Invalid HTTP reply (expected 200 got 500), make sure disco master is running at http://localhost:8989

 

Any idea? Thanks,

 

- Yan

 

 

[Ville Tuulos]

Hi Yan,

On Dec 7, 4:21 am, <Yan...@nokia.com> wrote:
> Hi,
>
> First, I followedhttp://discoproject.org/doc/start/install.html#setuptrying to set up a three-node cluster. It did not work when I tried to add nodes in the web interface. Then I followedhttp://discoproject.org/doc/start/troubleshoot.htmlto add nodes like [["localhost", "1"]] manually to disco_4441.config. But I still experienced some problems in the first paragraph in step 3) when I tried to run the sample job. The trace is copied below:
>
> disco@jackalope:~/disco-0.2.3/examples$ python count_words.pyhttp://localhost:8989
> Starting Disco job..
> Go tohttp://localhost:8989to see status of the job.

> Traceback (most recent call last):
>   File "count_words.py", line 22, in <module>
>     reduce = fun_reduce).wait()
>   File "/usr/local/lib/python2.6/dist-packages/disco-0.1-py2.6.egg/disco/core.py", line 225, in g
>     return f(*tuple([self.name] + list(args)), **kw)
>   File "/usr/local/lib/python2.6/dist-packages/disco-0.1-py2.6.egg/disco/core.py", line 154, in wait
>     status, results = self.results(name, timeout = poll_interval * 1000)
>   File "/usr/local/lib/python2.6/dist-packages/disco-0.1-py2.6.egg/disco/core.py", line 132, in results
>     results      = json.loads(self.request("/disco/ctrl/get_results", data))
>   File "/usr/local/lib/python2.6/dist-packages/disco-0.1-py2.6.egg/disco/core.py", line 54, in request
>     raise DiscoError('Got %s, make sure disco master is running at %s' % (e, self.host))

> disco.error.DiscoError: Got HTTP exception (http://localhost:8989/disco/ctrl/get_results):Downloadinghttp://localhost:8989/disco/ctrl/get_resultsfailed after 10 attempts: HTTP exception (http://localhost:8989/disco/ctrl/get_results):Invalid HTTP reply (expected 200 got 500), make sure disco master is running athttp://localhost:8989

Based on the stack trace above it seems that the job is started ok but
it fails for some reason.

Can you see the job on the web ui at http://localhost:8989? If you can
get to the job status page, do you see any errors / stack traces
there? Alternatively you can set the environment variable
DISCO_EVENTS=1 before running "python count_words.py" to get possible
errors printed out on the console. The error report produced by the
job should reveal why it fails.

Ville