-
Notifications
You must be signed in to change notification settings - Fork 128
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Do not write to the cache anywhere in build_target() #236
Comments
I recall at one point the value was returned using future (as well as written to disk) and the process was horribly slow and memory hungry for the host. For SLURM, this (at least) doubles the I/O writing; once to bring the data back to the host process and once to write it to disk from the host process. This will be a lot of work for the host process for a large project which usually has limited resources.
Ref: #115 |
So then we should think about a user-side option to cache inside or outside the future. Either way, the internals need to be cleaned up to account for all this. |
I wonder if there's a future environment variable that indicates whether the future has access to the main disk? That might be able to shield the user from yet another argument to make |
There may be, but I think there's also more to it than that: certain |
I think we're good here: in #237 and #227, all we need to do is pay attention to where we use |
drake_build()
should just return a target's value, and it should not cache errors or progress on its own. This is especially important for jobs deployed to remote locations that have no access to the cache. Transferring data should be the responsibility of thefuture
package.The text was updated successfully, but these errors were encountered: