I would like to run a master Julia process on a laptop with Mac OS (ideally as a jupyter notebook) and have the work done by remote processes on a cluster of Linux nodes. Or, to be more generic, to have a notebook in the browser on my local machine with the cluster as the computing backend.
I am able successfully add remote workers by calling
addprocs() on the laptop as follows:
addprocs([("email@example.com", 4)]; tunnel=true, dir="/home/ilyak/dpa-tools-benchmarking/testcase01", exename="julia", exeflags="--depwarn=no")
But then when I start Ioading modules, the module loading fails:
Deserialization checks failed while attempting to load cache from /Users/ilya/.julia/lib/v0.6/ProgressMeter.ji.
On the cluster, the Julia home dir is of course different (“/home/ilyak/.julia”)
Is there any way to solve this other than running the master process with exactly the same Julia build and install path?
Julia version is 0.6.0. On 0.5.2 did not work either, though with a different error message.
The exact experiment description and the code are at https://github.com/ikizhvatov/jlsca-tutorials/blob/master/HPC.md.