Fix a problem and uncover a new one...
May. 14th, 2010 07:34 pm![[personal profile]](https://www.dreamwidth.org/img/silk/identity/user.png){kind=small}
sawylI eventually managed to fix my UM compilation problem. When I checked the batch job used to drive the compile, I noticed that it included a LoadLeveler directive which explicitly set the shell to
But despite fixing this problem the model still failed to run correctly, crashing almost immediately after the start of the atmosphere model. Checking the output I found a whole load of errors from a script that appeared to be setting up the OASIS coupler — these might, for all I know, be perfectly normal — followed by a fatal error complaining of an atmosphere basis time mismatch. I'm not quite sure why this happened — I think the error indicates that the start dump does not match the model start time — because I simply copied the job from a colleague. It was at this point that the obvious occurred to me: I should simply get in contact with my colleague and get him to run the job. Far easier.
/bin/ksh. This overrode my normal shell, bash, and prevented my normal startup scripts, with their code to set up FCM, from being picked up. Once I knew this, I was quickly able to fix the problem by creating a .profile script which included a sensible basic path and enough of an environment to drive the build system.But despite fixing this problem the model still failed to run correctly, crashing almost immediately after the start of the atmosphere model. Checking the output I found a whole load of errors from a script that appeared to be setting up the OASIS coupler — these might, for all I know, be perfectly normal — followed by a fatal error complaining of an atmosphere basis time mismatch. I'm not quite sure why this happened — I think the error indicates that the start dump does not match the model start time — because I simply copied the job from a colleague. It was at this point that the obvious occurred to me: I should simply get in contact with my colleague and get him to run the job. Far easier.
