Running GEOS 5 on a GMAO Desktop: Difference between revisions

(6 intermediate revisions by 2 users not shown)

Line 1:

This page will detail how to run the GEOS 5 model on a GMAO desktop. Both the single column model and the full model at 2 degrees (144x91 or c48) can be run.

This page will detail how to run the GEOS 5 model on a GMAO desktop. Both the single column model and the full model at 2 degrees (144x91 or c48) can be run starting with Ganymed-4_0

==Prerequisites==

You will need a desktop running a 64 bit version of Linux. In addition you will need access to /ford1/local and /ford1/share from your desktop. GMAO support can help you with these items. If you want to run the full model you will need one of the newer 8 core (with hyperthreading) machines with 8 GB of memory. The single column model can be run on one of the older machines with less cores and memory.

You will need a desktop running a 64 bit version of Linux. In addition you will need access to /ford1/local and /ford1/share from your desktop. The ford1 disk contains the libraries, baselibs, and boundary conditions needed to run the model. GMAO support can help you with these two items. If you want to run the full model you will need one of the newer 8 core (with hyperthreading) machines with 8 GB of memory. The single column model can be run on one of the older machines with less cores and memory.

==Checking Out and Building==

Once you have satisfied the prerequisites you can check out and build the model. ~~First~~ you will need to set up a tunnel to the cvsacl server to access the repository. NCCS has a detailed description of how to do this at https://progress.nccs.nasa.gov/trac/admin/wiki/CVSACL.

Once you have satisfied the prerequisites you can check out and build the model. If you have not done so already you will need to upload the key for your local machine to progress. Then you will need to set up a tunnel to the cvsacl server to access the repository. NCCS has a detailed description of how to do this at https://progress.nccs.nasa.gov/trac/admin/wiki/CVSACL.

~~In brief~~ you must add something like this to your .ssh/config file:

You can upload your key at https://progress.nccs.nasa.gov/keyupload. You should be careful to concatenate all of your keys for the sites you access cvs from as everytime you upload a key as it wipes out your previous key file. For example if you are already accessing cvs from a different machine, copy the key from that machine to your desktop, put that in a file with your desktop key, and upload that file that contains both keys.

Next you must add something like this to your .ssh/config file:

<pre>

Host cvsacl

Line 21:

Line 22:

$ ssh cvsacl

</pre>

As long as this terminal is active you will be able to perform cvs commands and access the repository. Don't forget to set your CVSROOT environment variable. It should look like this:

If you get denied access with a keyboard-interactive error contact NCCS as you may not have an account on progress and cvsacl even if you are using cvsacldirect on Discover. As long as this terminal is active you will be able to perform cvs commands and access the repository. Don't forget to set your CVSROOT environment variable. It should look like this:

<pre>

:ext:user_name@ctunnel:/cvsroot/esma

Line 29:

Line 30:

$ cvs co -r Ganymed-4_0_BETA3 Ganymed

</pre>

You then need to update a few files. Issue to following command in the src directory to update these 4 files

<pre>

$ cvs upd -r mat-G40B3-desktop g5_modules Applications/GEOSgcm_App/gcm_convert.j Applications/GEOSgcm_App/gcm_setup Applications/GEOSgcm_App/scm_setup

</pre>

Once this is done you can build the model as usual. Make sure that the environment variable ESMADIR is set, this will be path_model/GEOSagcm. Then

<pre>

$ source $ESMADIR/src/g5_modules.

</pre>

and issue the make. You can and should use the parallel build up to the number of cores you have available. To build using 4 cores you would issue:

<pre>

$ make -j4 pinstall

</pre>

Assuming everything goes right you should have a functional GEOS 5 model.

==Running the Model==

You can setup an experiment now as usual. To do a full model run use the gcm_setup script as usual. Once the experiment is setup and you have your restarts copied run the gcm_run.j script. To run the single column model copy the scm_setup script to the directory you want to run the experiment in. Run the scm_setup script in this directory. Currently on /ford1 are boundary conditions for 144x91 and c48 using the Reynolds ocean. If you have a newer machine with 8 cores (including hyperthreading) you can run 144x91 on a 2x4 layout and c48 on a 1x6 layout. For reference, using c48, on the Reynolds ocean, no gocart, with the standard history had a highwater memory use of 7.3 GB and took about 9 minutes running on a 1x6 layout.

@@ Line 1: / Line 1: @@
-This page will detail how to run the GEOS 5 model on a GMAO desktop. Both the single column model and the full model at 2 degrees (144x91 or c48) can be run.
+This page will detail how to run the GEOS 5 model on a GMAO desktop. Both the single column model and the full model at 2 degrees (144x91 or c48) can be run starting with Ganymed-4_0
 ==Prerequisites==
-You will need a desktop running a 64 bit version of Linux. In addition you will need access to /ford1/local and /ford1/share from your desktop. GMAO support can help you with these items. If you want to run the full model you will need one of the newer 8 core (with hyperthreading) machines with 8 GB of memory. The single column model can be run on one of the older machines with less cores and memory.
+You will need a desktop running a 64 bit version of Linux. In addition you will need access to /ford1/local and /ford1/share from your desktop. The ford1 disk contains the libraries, baselibs, and boundary conditions needed to run the model. GMAO support can help you with these two items. If you want to run the full model you will need one of the newer 8 core (with hyperthreading) machines with 8 GB of memory. The single column model can be run on one of the older machines with less cores and memory.
 ==Checking Out and Building==
-Once you have satisfied the prerequisites you can check out and build the model. First you will need to set up a tunnel to the cvsacl server to access the repository. NCCS has a detailed description of how to do this at https://progress.nccs.nasa.gov/trac/admin/wiki/CVSACL.
+Once you have satisfied the prerequisites you can check out and build the model. If you have not done so already you will need to upload the key for your local machine to progress. Then you will need to set up a tunnel to the cvsacl server to access the repository. NCCS has a detailed description of how to do this at https://progress.nccs.nasa.gov/trac/admin/wiki/CVSACL.
-In brief you must add something like this to your .ssh/config file:
+You can upload your key at https://progress.nccs.nasa.gov/keyupload. You should be careful to concatenate all of your keys for the sites you access cvs from as everytime you upload a key as it wipes out your previous key file. For example if you are already accessing cvs from a different machine, copy the key from that machine to your desktop, put that in a file with your desktop key, and upload that file that contains both keys.
+Next you must add something like this to your .ssh/config file:
 <pre>
 Host cvsacl
@@ Line 21: / Line 22: @@
 $ ssh cvsacl
 </pre>
-As long as this terminal is active you will be able to perform cvs commands and access the repository. Don't forget to set your CVSROOT environment variable. It should look like this:
+If you get denied access with a keyboard-interactive error contact NCCS as you may not have an account on progress and cvsacl even if you are using cvsacldirect on Discover. As long as this terminal is active you will be able to perform cvs commands and access the repository. Don't forget to set your CVSROOT environment variable. It should look like this:
 <pre>
 :ext:user_name@ctunnel:/cvsroot/esma
@@ Line 29: / Line 30: @@
 $ cvs co -r Ganymed-4_0_BETA3 Ganymed
 </pre>
+You then need to update a few files. Issue to following command in the src directory to update these 4 files
+<pre>
+$ cvs upd -r mat-G40B3-desktop g5_modules Applications/GEOSgcm_App/gcm_convert.j Applications/GEOSgcm_App/gcm_setup Applications/GEOSgcm_App/scm_setup
+</pre>
+Once this is done you can build the model as usual. Make sure that the environment variable ESMADIR is set, this will be path_model/GEOSagcm. Then
+<pre>
+$ source $ESMADIR/src/g5_modules.
+</pre>
+and issue the make. You can and should use the parallel build up to the number of cores you have available. To build using 4 cores you would issue:
+<pre>
+$ make -j4 pinstall
+</pre>
+Assuming everything goes right you should have a functional GEOS 5 model.
+==Running the Model==
+You can setup an experiment now as usual. To do a full model run use the gcm_setup script as usual. Once the experiment is setup and you have your restarts copied run the gcm_run.j script. To run the single column model copy the scm_setup script to the directory you want to run the experiment in. Run the scm_setup script in this directory. Currently on /ford1 are boundary conditions for 144x91 and c48 using the Reynolds ocean. If you have a newer machine with 8 cores (including hyperthreading) you can run 144x91 on a 2x4 layout and c48 on a 1x6 layout. For reference, using c48, on the Reynolds ocean, no gocart, with the standard history had a highwater memory use of 7.3 GB and took about 9 minutes running on a 1x6 layout.