Grid Integration Tips
If you are having problems with your CLC Grid Integration, please check the following points:
- Does your system meets the Requirements for CLC Grid Integration? For example, please check that the machine the CLC Science Server is running on is configured as a submit host for your grid system, and please check that you are running Sun/Oracle Java 1.7 on all execution hosts.
- The user running the CLC Science Server process is the same user seen as the submitter of all jobs to the grid. Does this user exist on your grid nodes? Does it have permission to submit to the necessary queues, and to write to the shared directories identified in the Grid Preset(s) and any
clcgridworker.vmoptionsfiles? - Are your CLC Science Server file locations mounted with the same path on the grid nodes as on the master CLC Science Server and accessible to the user that runs the CLC Science Server process?
- If you store data in a database, are all the grid nodes able to access that database, using the user account of the user running the CLC Science Server process?
- If you store data in a database, did you enter a machine name in the Host box of the Database Location field when seeting up the Database Location using the CLC Science Server web administration form? In particular, a generic term such as
localhostwill not work, as the grid nodes will not be able to find the host with the database on it using that information. - If you installed the CLC Science Server as root, and then later decided to run it as a non-privileged user, please ensure that you stop the server, recursively change ownership on the CLC Science Server installation directory and any data locations assigned to the CLC Server. Please restart the server as the new user. You may need to Rebuilding the index after you restart the server.
- Is your java binary on the PATH? If not, then either add it to PATH, or edit the clcgridworker script in the CLC Science Server installation area, with the relative path from this location:
gridres/dist/clcgridworker, and set the JAVA variable to the full path of your java binary. Then re-save each of your grid presets, so that this altered clcgridworker script is deployed to the location specified in the Path to CLC Grid Worker field of your preset. - Is the
SGE_ROOTvariable set early enough in your system that it is included in the environment of services? Alternatively, did you edit the CLC Science Serverr startup script to set this variable? If so, the script is overwritten on upgrade - you will need to re-add this variable setting, either to the startup script, or system wide in such a way that it is available in the environment of your services. - Is your java 64-bit, while your DRMAA library is 32-bit, or vice versa? These two things must be either both for 64-bit systems or both for 32-bit systems.