On Mon, May 6, 2013 at 9:35 AM, Krampis, Konstantinos <kkrampis@jcvi.org> wrote:
Hi John,
thanks for the quick reply! It is actually a virtual cluster using virtual machines with VirtualBox, which are brought up via Vagrant (www.vagrantup.com) - that also sets up the shared filesystem. I chose LWR for the reason that it not a real cluster, and needed something lightweight instead of having to stand up SGE etc.
Fascinating. Unless you don't have root on the physical server, I would still really recommend SGE. I cannot see any reason it shouldn't work on virtual servers, it works with CloudMan and there the master and workers are both virtualized.
I am very interested in the 3rd option you present in your response below. I am still confused though that if Galaxy and LWR doe not use shared filesystem, how does Galaxy know when the job is finished, or where to find the output files ?
Or does the output gets "piped" back to Galaxy ? That would work (all virtual machines are on the same box, so no latency), as long as the output gets deleted by LWR after being send to Galaxy (to save space in the worker virtual machines).
The LWR job runner communicates with the LWR client over HTTP to work out these details. The job description is rewritten with new file paths (input, output, tool files, config files), staged and submitted to the remote LWR server which executes it and tracks its progress. Meanwhile, the LWR job runner polls the LWR server waiting for completion of the job. Upon completion, it downloads the outputs and places them in the paths Galaxy is expecting. This is quite different than the other job runners, so there are all sorts of assumptions tools may make that could cause this to fail - so if you are going to run more than a handful of tools this way I would really recommend the SGE or otherwise you are going to spend time tracking down little annoying problems I suspect. As for the option of not staging the files, I am going to make the rookie mistake of suggesting that this is straight forward to implement in your case - as long as all of the files are available on the remote system. I have created an issue for this, I don't know when I will get to this but you can subscribe to the issue to stay in the loop: https://bitbucket.org/jmchilton/lwr/issue/8/allow-alternate-staging-rules To outline how to just disable staging, I think most of the changes will need to be made to
If you could provide some pointers to documentation / code, or explain how could be set up it would be of great help !
many thanks,
Ntino
________________________________________ From: jmchilton@gmail.com [jmchilton@gmail.com] On Behalf Of John Chilton [chilton@msi.umn.edu] Sent: Monday, May 06, 2013 9:50 AM To: Krampis, Konstantinos Cc: galaxy-dev@lists.bx.psu.edu Subject: Re: [galaxy-dev] LWR runner configuration for shared folder in cluster
Hello Dr. Krampis,
At the present, the LWR is most valuable when there is not a shared file system between the server executing the jobs and the server hosting Galaxy. In this case you seem to have a shared filesystem so I would think setting up something like sun grid engine to manage the jobs and using the DRMAA job runner would be the best route forward.
The LWR and the corresponding galaxy job runner will coordinate to stage jobs, but the upshot is that the LWR should be the only thing writing to its staging directory. In this case you have configured the LWR and Galaxy to both use the same directory. You should change this configuration immediately, I am worried the LWR is going to delete or overwrite files maintained by Galaxy. I am sorry for the confusion, I will update the documentation to explicitly warn against this.
If you still feel there is a compelling reason to use the LWR in this situation, you will just want to change the staging_directory in the LWR configuration to something else. It has long been on my TODO list to allow one to disable (or selectively disable by path regex/globs) file staging with the LWR, it seems like that would what would also help in your situation. Let me know if that is of interest to you.
-John
On Mon, May 6, 2013 at 8:32 AM, Krampis, Konstantinos <kkrampis@jcvi.org> wrote:
Hi all,
I am trying to set up a Galaxy cluster using the LWR runner. The nodes have a shared filesystem and in universe.wsgi this parameter is set :
job_working_directory = /mnt/shared ... clustalw = lwr://http://192.168.33.12:8913 ....
this folder has been "chown-ed" to the galaxy user, and also is "a+w", while it has been verified that can been read / written by ssh-ing to each node of the cluster. The sticky bit is set.
When I try to run jobs (I used clustalw as example) there seems to be confusion between where Galaxy puts files and where LWR tries to read them from. Here are two setups that error out:
1). When in server.ini for LWR the following is set as: staging_directory = /mnt/shared/000
galaxy error:
galaxy.jobs DEBUG 2013-05-06 10:21:22,320 (128) Working directory for job is: /mnt/shared/000/128 galaxy.jobs.handler DEBUG 2013-05-06 10:21:22,320 dispatching job 128 to lwr runner galaxy.jobs.handler INFO 2013-05-06 10:21:22,427 (128) Job dispatched galaxy.datatypes.metadata DEBUG 2013-05-06 10:21:22,875 Cleaning up external metadata files galaxy.jobs.runners.lwr ERROR 2013-05-06 10:21:22,902 failure running job 128
lwr error (on the cluster node):
File "/home/vagrant/jmchilton-lwr-5213f6dce32d/lwr/app.py", line 81, in setup manager.setup_job_directory(job_id) File "/home/vagrant/jmchilton-lwr-5213f6dce32d/lwr/manager.py", line 101, in setup_job_directory os.mkdir(job_directory) OSError: [Errno 17] File exists: '/mnt/shared/000/128'
2). When in server.ini for LWR the following is set as: staging_directory = /mnt/shared
galaxy error:
galaxy.jobs DEBUG 2013-05-06 10:28:46,872 (129) Working directory for job is: /mnt/shared/000/129 galaxy.jobs.handler DEBUG 2013-05-06 10:28:46,872 dispatching job 129 to lwr runner galaxy.jobs.handler INFO 2013-05-06 10:28:46,967 (129) Job dispatched 192.168.33.1 - - [06/May/2013:10:28:48 -0200] "GET /api/histories/2a56795cad3c7db3 HTTP/1.1" 200 - "http://192.168.33.11:8080/history" "Mozilla/5.0 (Macintosh; Intel Mac OS X 10_7_5) AppleWebKit/537.31 (KHTML, like Gecko) Chrome/26.0.1410.65 Safari/537.31" galaxy.jobs.runners.lwr DEBUG 2013-05-06 10:28:50,653 run_results {'status': 'status', 'returncode': 0, 'complete': 'true', 'stderr': '', 'stdout': ''} galaxy.datatypes.metadata DEBUG 2013-05-06 10:28:50,970 Cleaning up external metadata files galaxy.jobs.runners.lwr ERROR 2013-05-06 10:28:51,050 failure running job 129
lwr error (on the cluster node):
resp.app_iter = FileIterator(result) File "/home/vagrant/jmchilton-lwr-5213f6dce32d/lwr/framework.py", line 111, in __init__ self.input = open(path, 'rb') IOError: [Errno 2] No such file or directory: u'/mnt/shared/129/outputs/dataset_170.dat'
The full error stacks are at the end of this email. It might be something very simple that I am missing, but any feedback would be greatly appreciated. Thanks !
Ntino
-- Konstantinos (Ntino) Krampis, Ph.D. Asst. Professor, Informatics J.Craig Venter Institute
kkrampis@jcvi.org agbiotec@gmail.com +1-540-200-8277
Web: http://bit.ly/cloud-research http://cloudbiolinux.org/ http://twitter.com/agbiotec
---- GALAXY ERROR
galaxy.jobs DEBUG 2013-05-06 10:21:22,320 (128) Working directory for job is: /mnt/shared/000/128 galaxy.jobs.handler DEBUG 2013-05-06 10:21:22,320 dispatching job 128 to lwr runner galaxy.jobs.handler INFO 2013-05-06 10:21:22,427 (128) Job dispatched galaxy.datatypes.metadata DEBUG 2013-05-06 10:21:22,875 Cleaning up external metadata files galaxy.jobs.runners.lwr ERROR 2013-05-06 10:21:22,902 failure running job 128 Traceback (most recent call last): File "/home/vagrant/galaxy-dist/lib/galaxy/jobs/runners/lwr.py", line 286, in run_job file_stager = FileStager(client, command_line, job_wrapper.extra_filenames, input_files, output_files, job_wrapper.tool.tool_dir) File "/home/vagrant/galaxy-dist/lib/galaxy/jobs/runners/lwr.py", line 40, in __init__ job_config = client.setup() File "/home/vagrant/galaxy-dist/lib/galaxy/jobs/runners/lwr.py", line 212, in setup return self.__raw_execute_and_parse("setup", { "job_id" : self.job_id }) File "/home/vagrant/galaxy-dist/lib/galaxy/jobs/runners/lwr.py", line 150, in __raw_execute_and_parse response = self.__raw_execute(command, args, data) File "/home/vagrant/galaxy-dist/lib/galaxy/jobs/runners/lwr.py", line 146, in __raw_execute response = self.url_open(request, data) File "/home/vagrant/galaxy-dist/lib/galaxy/jobs/runners/lwr.py", line 134, in url_open return urllib2.urlopen(request, data) File "/usr/lib/python2.7/urllib2.py", line 126, in urlopen return _opener.open(url, data, timeout) File "/usr/lib/python2.7/urllib2.py", line 406, in open response = meth(req, response) File "/usr/lib/python2.7/urllib2.py", line 519, in http_response 'http', request, response, code, msg, hdrs) File "/usr/lib/python2.7/urllib2.py", line 444, in error return self._call_chain(*args) File "/usr/lib/python2.7/urllib2.py", line 378, in _call_chain result = func(*args) File "/usr/lib/python2.7/urllib2.py", line 527, in http_error_default raise HTTPError(req.get_full_url(), code, msg, hdrs, fp) HTTPError: HTTP Error 500: Internal Server Error
---- LWR ERROR
Exception happened during processing of request from ('192.168.33.11', 44802) Traceback (most recent call last): File "/usr/local/lib/python2.7/dist-packages/Paste-1.7.5.1-py2.7.egg/paste/httpserver.py", line 1068, in process_request_in_thread self.finish_request(request, client_address) File "/usr/lib/python2.7/SocketServer.py", line 323, in finish_request self.RequestHandlerClass(request, client_address, self) File "/usr/lib/python2.7/SocketServer.py", line 638, in __init__ self.handle() File "/usr/local/lib/python2.7/dist-packages/Paste-1.7.5.1-py2.7.egg/paste/httpserver.py", line 442, in handle BaseHTTPRequestHandler.handle(self) File "/usr/lib/python2.7/BaseHTTPServer.py", line 340, in handle self.handle_one_request() File "/usr/local/lib/python2.7/dist-packages/Paste-1.7.5.1-py2.7.egg/paste/httpserver.py", line 437, in handle_one_request self.wsgi_execute() File "/usr/local/lib/python2.7/dist-packages/Paste-1.7.5.1-py2.7.egg/paste/httpserver.py", line 287, in wsgi_execute self.wsgi_start_response) File "/home/vagrant/jmchilton-lwr-5213f6dce32d/lwr/framework.py", line 35, in __call__ return controller(environ, start_response, **request_args) File "/home/vagrant/jmchilton-lwr-5213f6dce32d/lwr/framework.py", line 90, in controller_replacement result = func(**args) File "/home/vagrant/jmchilton-lwr-5213f6dce32d/lwr/app.py", line 81, in setup manager.setup_job_directory(job_id) File "/home/vagrant/jmchilton-lwr-5213f6dce32d/lwr/manager.py", line 101, in setup_job_directory os.mkdir(job_directory) OSError: [Errno 17] File exists: '/mnt/shared/000/128'
___________________________________________________________ Please keep all replies on the list by using "reply all" in your mail client. To manage your subscriptions to this and other Galaxy lists, please use the interface at: http://lists.bx.psu.edu/
To search Galaxy mailing lists use the unified search at: http://galaxyproject.org/search/mailinglists/