Re: [galaxy-dev] LD_LIBRARY_PATH not exporting to galaxy
Since there are so many surprising new details in your last mail, maybe it would help if you could tell where you got your Python from. Is it a regular system installation or a toolshed installed version inside Galaxy (which may use its own env.sh to manipulate paths) ? Best, Wolfgang On 29.01.2015 11:02, Nikhil Joshi wrote: Ok, a little more progress. This is really strange. It seems that it is only python that is not getting LD_LIBRARY_PATH, and it only doesn't get it if it is being invoked by a user (i.e. not root). If root invokes python then it gets the variable no problem. If the galaxy user invokes it, it does not get it. However, if the galaxy user invokes perl and prints LD_LIBRARY_PATH, it is there?!?!? I should mention that this Galaxy is being run on an Amazon AMI. Is there some weird security thing where *specifically* python does not get LD_LIBRARY_PATH? - Nik. On Wed, Jan 28, 2015 at 4:33 PM, Nikhil Joshi najo...@ucdavis.edu mailto:najo...@ucdavis.edu wrote: Well we are using the environment_setup_file option in our galaxy.ini file. Supposedly we can point to a shell script that will be sourced before running each job. In this script we load all of the environment variables using modules. The paths are all correct after this runs. However, the python script that is run (from the XML for the cufflinks tool) does not inherit these environment variables correctly. Below is the content of our env.sh script: #!/bin/bash . /etc/profile.d/modules_sh.sh module load meta/galaxy This definitely loads the environment properly. And here is the cufflinks line from the XML: command interpreter=python cufflinks_wrapper.py --input=$input . etc. This cufflinks wrapper fails because it cannot actually run cufflinks because it cannot find a library that it needs, because the LD_LIBRARY_PATH is empty within its environment. I actually edited the XML to echo the load path to a file before the python script runs and the path was fine, but still the python script doesn't inherit it properly. - Nik. On Wed, Jan 28, 2015 at 6:16 AM, Nate Coraor n...@bx.psu.edu mailto:n...@bx.psu.edu wrote: Hi Nik, Is there any chance that $LD_LIBRARY_PATH is being set in the parent shell but not exported? --nate On Wed, Jan 28, 2015 at 7:04 AM, Nikhil Joshi najo...@ucdavis.edu mailto:najo...@ucdavis.edu wrote: To answer your questions from above, in case it can help with this new problem, we are sourcing the modules from an env.sh script whose full path is the value of the environment_setup_file option. This fixed our first problem, but not this second one. We created the modules ourselves for each of the programs. - Nik. On Tue, Jan 27, 2015 at 6:45 PM, Nikhil Joshi najo...@ucdavis.edu mailto:najo...@ucdavis.edu wrote: Okay, we fixed that problem, but another tool is now having the same problem. Now I am trying to run cufflinks and I am getting that same error, but this time there is yet another layer of abstraction because the XML is calling a python script. It seems like the LD_LIBRARY_PATH is correct right up until the python wrapper script gets called and then for some reason the environment is not being passed to the script. This seems like a problem with Popen somewhere in galaxy where it is not passing the environment to the subprocess? - Nik. On Tue, Jan 27, 2015 at 4:33 PM, Nikhil Joshi najo...@ucdavis.edu mailto:najo...@ucdavis.edu wrote: Hi John, Thanks for your reply. Our sys admin figured out the problem we had to use the environment_setup_file option to load the modules and we had to fix some shell issues where galaxy was using the dash shell instead of bash. Anyways, it's fixed for now! - Nik. On Mon, Jan 26, 2015 at 6:43 AM, John Chilton jmchil...@gmail.com mailto:jmchil...@gmail.com wrote: Had some idea but they were all dead ends. How are you sourcing the modules - in Galaxy's environment, using an environment file, using Galaxy dependencies (env.sh files) or using Galaxy's modules support? Do you have an example of such a module that
Re: [galaxy-dev] LD_LIBRARY_PATH not exporting to galaxy
It is a regular system installation. Just for a sanity check, I checked the md5 hash with the same version of my local python and it is the same. Also, as a sanity check, I ran the tests on my local machine and python got the LD_LIBRARY_PATH variable just fine, both as a regular user and as root. I thought that maybe it was an apparmor thing, but usually that logs when it denies an application, and I didn't find anything in the logs. - Nik. On Thu, Jan 29, 2015 at 5:02 AM, Wolfgang Maier wolfgang.ma...@biologie.uni-freiburg.de wrote: Since there are so many surprising new details in your last mail, maybe it would help if you could tell where you got your Python from. Is it a regular system installation or a toolshed installed version inside Galaxy (which may use its own env.sh to manipulate paths) ? Best, Wolfgang On 29.01.2015 11:02, Nikhil Joshi wrote: Ok, a little more progress. This is really strange. It seems that it is only python that is not getting LD_LIBRARY_PATH, and it only doesn't get it if it is being invoked by a user (i.e. not root). If root invokes python then it gets the variable no problem. If the galaxy user invokes it, it does not get it. However, if the galaxy user invokes perl and prints LD_LIBRARY_PATH, it is there?!?!? I should mention that this Galaxy is being run on an Amazon AMI. Is there some weird security thing where *specifically* python does not get LD_LIBRARY_PATH? - Nik. On Wed, Jan 28, 2015 at 4:33 PM, Nikhil Joshi najo...@ucdavis.edu mailto:najo...@ucdavis.edu wrote: Well we are using the environment_setup_file option in our galaxy.ini file. Supposedly we can point to a shell script that will be sourced before running each job. In this script we load all of the environment variables using modules. The paths are all correct after this runs. However, the python script that is run (from the XML for the cufflinks tool) does not inherit these environment variables correctly. Below is the content of our env.sh script: #!/bin/bash . /etc/profile.d/modules_sh.sh module load meta/galaxy This definitely loads the environment properly. And here is the cufflinks line from the XML: command interpreter=python cufflinks_wrapper.py --input=$input . etc. This cufflinks wrapper fails because it cannot actually run cufflinks because it cannot find a library that it needs, because the LD_LIBRARY_PATH is empty within its environment. I actually edited the XML to echo the load path to a file before the python script runs and the path was fine, but still the python script doesn't inherit it properly. - Nik. On Wed, Jan 28, 2015 at 6:16 AM, Nate Coraor n...@bx.psu.edu mailto:n...@bx.psu.edu wrote: Hi Nik, Is there any chance that $LD_LIBRARY_PATH is being set in the parent shell but not exported? --nate On Wed, Jan 28, 2015 at 7:04 AM, Nikhil Joshi najo...@ucdavis.edu mailto:najo...@ucdavis.edu wrote: To answer your questions from above, in case it can help with this new problem, we are sourcing the modules from an env.sh script whose full path is the value of the environment_setup_file option. This fixed our first problem, but not this second one. We created the modules ourselves for each of the programs. - Nik. On Tue, Jan 27, 2015 at 6:45 PM, Nikhil Joshi najo...@ucdavis.edu mailto:najo...@ucdavis.edu wrote: Okay, we fixed that problem, but another tool is now having the same problem. Now I am trying to run cufflinks and I am getting that same error, but this time there is yet another layer of abstraction because the XML is calling a python script. It seems like the LD_LIBRARY_PATH is correct right up until the python wrapper script gets called and then for some reason the environment is not being passed to the script. This seems like a problem with Popen somewhere in galaxy where it is not passing the environment to the subprocess? - Nik. On Tue, Jan 27, 2015 at 4:33 PM, Nikhil Joshi najo...@ucdavis.edu mailto:najo...@ucdavis.edu wrote: Hi John, Thanks for your reply. Our sys admin figured out the problem we had to use the environment_setup_file option to load the modules and we had to fix some shell issues where galaxy was using the dash shell instead of bash. Anyways, it's fixed for now!
Re: [galaxy-dev] LD_LIBRARY_PATH not exporting to galaxy
Well we are using the environment_setup_file option in our galaxy.ini file. Supposedly we can point to a shell script that will be sourced before running each job. In this script we load all of the environment variables using modules. The paths are all correct after this runs. However, the python script that is run (from the XML for the cufflinks tool) does not inherit these environment variables correctly. Below is the content of our env.sh script: #!/bin/bash . /etc/profile.d/modules_sh.sh module load meta/galaxy This definitely loads the environment properly. And here is the cufflinks line from the XML: command interpreter=python cufflinks_wrapper.py --input=$input . etc. This cufflinks wrapper fails because it cannot actually run cufflinks because it cannot find a library that it needs, because the LD_LIBRARY_PATH is empty within its environment. I actually edited the XML to echo the load path to a file before the python script runs and the path was fine, but still the python script doesn't inherit it properly. - Nik. On Wed, Jan 28, 2015 at 6:16 AM, Nate Coraor n...@bx.psu.edu wrote: Hi Nik, Is there any chance that $LD_LIBRARY_PATH is being set in the parent shell but not exported? --nate On Wed, Jan 28, 2015 at 7:04 AM, Nikhil Joshi najo...@ucdavis.edu wrote: To answer your questions from above, in case it can help with this new problem, we are sourcing the modules from an env.sh script whose full path is the value of the environment_setup_file option. This fixed our first problem, but not this second one. We created the modules ourselves for each of the programs. - Nik. On Tue, Jan 27, 2015 at 6:45 PM, Nikhil Joshi najo...@ucdavis.edu wrote: Okay, we fixed that problem, but another tool is now having the same problem. Now I am trying to run cufflinks and I am getting that same error, but this time there is yet another layer of abstraction because the XML is calling a python script. It seems like the LD_LIBRARY_PATH is correct right up until the python wrapper script gets called and then for some reason the environment is not being passed to the script. This seems like a problem with Popen somewhere in galaxy where it is not passing the environment to the subprocess? - Nik. On Tue, Jan 27, 2015 at 4:33 PM, Nikhil Joshi najo...@ucdavis.edu wrote: Hi John, Thanks for your reply. Our sys admin figured out the problem we had to use the environment_setup_file option to load the modules and we had to fix some shell issues where galaxy was using the dash shell instead of bash. Anyways, it's fixed for now! - Nik. On Mon, Jan 26, 2015 at 6:43 AM, John Chilton jmchil...@gmail.com wrote: Had some idea but they were all dead ends. How are you sourcing the modules - in Galaxy's environment, using an environment file, using Galaxy dependencies (env.sh files) or using Galaxy's modules support? Do you have an example of such a module that sets LD_LIBRARY_PATH? One thing to check is the Galaxy's .bashrc file - make sure it isn't resetting LD_LIBRARY_PATH somehow. -John On Fri, Jan 23, 2015 at 9:34 PM, Nikhil Joshi najo...@ucdavis.edu wrote: Hi all, So we have forked the stable branch of galaxy and are making our own modifications. We are running galaxy on Ubuntu 14.04.1 LTS and we're using modules to load the underlying software, i.e. set environment variables. This works just fine for PATH, JAVA_JAR_PATH, PYTHONPATH, etc., however, LD_LIBRARY_PATH is not being passed to the tools. When a tool (such as cuffdiff) tries to access LD_LIBRARY_PATH, it is empty. We've checked to make sure that loading the cufflinks module correctly updates the LD_LIBRARY_PATH, however, when it is used in galaxy, all the other environment variables are set properly, but not LD_LIBRARY_PATH. Does anyone have any ideas on why this would happen and how to fix it? - Nik. -- Nikhil Joshi Bioinformatics Analyst/Programmer UC Davis Bioinformatics Core http://bioinformatics.ucdavis.edu/ najoshi -at- ucdavis -dot- edu 530.752.2698 (w) ___ 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: https://lists.galaxyproject.org/ To search Galaxy mailing lists use the unified search at: http://galaxyproject.org/search/mailinglists/ -- Nikhil Joshi Bioinformatics Analyst/Programmer UC Davis Bioinformatics Core http://bioinformatics.ucdavis.edu/ najoshi -at- ucdavis -dot- edu 530.752.2698 (w) -- Nikhil Joshi Bioinformatics Analyst/Programmer UC Davis Bioinformatics Core http://bioinformatics.ucdavis.edu/ najoshi -at- ucdavis -dot- edu 530.752.2698 (w) -- Nikhil Joshi Bioinformatics Analyst/Programmer UC Davis Bioinformatics Core http://bioinformatics.ucdavis.edu/ najoshi -at- ucdavis
