I have updated the option lists in Simfactory for several machines. This is to ensure that these option lists and machine descriptions are current, so that these machines can be used easily. I have tested that the test suites can be run, but not whether they succeed.
According to my tests, these option lists now allow syncing, building, submitting, and running on the following systems without problems: - Bethe, Blue Waters, Gordon, Hopper, Kraken, Lonestar, Mike, Nvidia, Orca, Philip, Queen Bee, Redshift, Stampede, Stampede-MIC, Titan, Trestles, Zwicky
There are problems on the following systems: - Carver, Pandora, Surveyor (job submission) - Vesta (running test suite)
I did not test the following systems: - Datura, GPC, Loewe, Supermuc
As mentioned above, I did not check the results of running the test cases; I assume that some test cases are failing on some of these systems.
I think we should address the problems on Pandora, Datura, and Vesta, and maybe Loewe. I volunteer for Vesta.
I am using the "distribute" script that comes with Simfactory for this. To use it: - Check out the Einstein Toolkit - run e.g. ./sim/bin/distribute --no-benchmark queenbee - this will sync, and then remotely build, submit a simple test job, run the test cases, and then test checkpoint/recovery/presubmission - high-level output: tail -f log/queenbee.results - verbose output: tail -f log/queenbee.out
Since this works remotely, this requires keeping a network connection from your machine to the machine being tested. I thus run this from a workstation, not from my laptop. If you run this on Queen Bee directly, you obviously don't need to keep a network connection since the tests will run locally.
Other important options for "distribute": - --no-clean: don't rebuild from scratch - --no-recover: don't test checkpoint/recovery; this step will fail without a queueing system (e.g. on a workstation)
-erik