Adding processes to HTCondor

[juliaparallel]

Adding processes to HTCondor

[grero]

Hi,
Did you try looking at the error produced by the worker? I faced a similar problem, and looking at the output files in ~/.julia-htc/julia--1.e I discovered that the environment was not exported, which gave an error like this:

fatal: error thrown and no exception handler available.
Base.InitError(mod=:Pkg, error=Base.KeyError(key="HOME"))

I'm still working on how to resolve that issue, though.

[EdgarMCR]

Hi,
I get the same error as grero but I am just using vanilla julia and condor and run the julia file from a bash script. At the start of the bash script I run
eval julia --version
which nicely outputs
julia version 0.5.0

However, running any julia file produces the same error as @grero.

[azraq27]

I just submitted a patch that fixes it on my machine. All it does it add the HOME environment variable to the Condor worker submission

[grero]

Just applied the patch from @azraq27 on my local clone of the repo. It seemed to help (I also had to add the PATH environment variable), in that the worker doesn't die immediately, but now it just seems to wait indefinitely to connect to the master process. This could be a network issue, but I'm not at all sure how to debug it, since no output is produced.

[azraq27]

It writes the STDOUT and STDERR into *.o and *.e files in the ~/.julia-htc directory. Does the STDERR give you some idea of why the process isn't starting?

[grero]

Thanks, it turns out that there was a problem with my network setup, unrelated to ClusterManagers. The patch completely solved the problem on the julia end for me.

[ivborissov]

Hi, does condor.jl support running HTCondor on Windows environment?

[juliohm]

Too old to reproduce.