Closed falkamelung closed 3 years ago
What's the actual error you get from Dask?
It does not give any error. The jobs start but don’t run. The jobs stop when the walltime is over.
The way to debug this may be to run one of the examples that Dask comes with and see whether it also fails with our environment.
reloved a while ago
I installed the python environments (new and old) in your area (test/test2 and test/testold). Some description of the issue is also at https://github.com/insarlab/MintPy/issues/165 . Since then somebody confirmed that it does not work under PBS either. David's effort are documented at https://github.com/2gotgrossman/dask-rsmas-presentation . As I said, I spent lots of time to install the old environment using the old requirements files but did not get it work.
_(After opening the issue it occurred to me that this is not a rsmasinsar but a MintPy issue. The MintPy environment is simpler as it does not have ISCE. I did install a mintpy python environment (run
s.bmintpy
). It gives the same problem @yunjunz is also interested in this.).First run test data with the old (good) python environment (in
3rdparty
dir usingln -s /projects/scratch/insarlab/famelung/MINICONDA3_GOOD miniconda3
):You will see the following output on the screen. Once you see the line 'FUTURE #1...` that means the first worker has completed its job.
To run the current (new) python environment (installed in
/3rparty
dir as described in https://github.com/geodesymiami/rsmas_insar/blob/master/docs/installation.md#installation-guide ) just do (after clearing your old environment) usingand the same commands above. You will see the screen output below, but the
FUTURE #1
will never show up. If you runbjobs
you will see that the worker have been started but the don't run. They stop after the time-out period of 30 minutes.