You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I have a julia package "packageName.jl" that has CondaPkg.jl as its dependency. When I now run "using packageName" while using MPI so on every rank, I get this a bunch of times:
Info: CondaPkg: Waiting for lock to be freed. You may delete this file if no other process is resolving.
And at the end, the program crashes, I presume because the lock could not be freed...
A workaround I came up with is to run "using packageName" in serial on each rank. However, for a large enough world size, this takes too long.
I did also load the package before starting the MPI job so that changes can be resolved. So I'm sure there are no changes that have to be accounted for when I start the MPI job.
Is there a way to skip this part so that I can load the package in parallel? Maybe with an unsafe flag?
I appreciate any help, even though my code is too complex to give a minimal example now.
Thank you in advance :)
The text was updated successfully, but these errors were encountered:
I have a julia package "packageName.jl" that has CondaPkg.jl as its dependency. When I now run "using packageName" while using MPI so on every rank, I get this a bunch of times:
Info: CondaPkg: Waiting for lock to be freed. You may delete this file if no other process is resolving.
And at the end, the program crashes, I presume because the lock could not be freed...
A workaround I came up with is to run "using packageName" in serial on each rank. However, for a large enough world size, this takes too long.
I did also load the package before starting the MPI job so that changes can be resolved. So I'm sure there are no changes that have to be accounted for when I start the MPI job.
Is there a way to skip this part so that I can load the package in parallel? Maybe with an unsafe flag?
I appreciate any help, even though my code is too complex to give a minimal example now.
Thank you in advance :)
The text was updated successfully, but these errors were encountered: