parallel get from datalad archive gives error
bpinsard opened this issue · comments
Basile commented
What is the problem?
running datalad get -J 8 .
in a dataset with a single datalad archive will give the following error:
[git-annex: .git/annex/tmp/MD5E-s8351481--f35cc3878536756c8567cc9d2421f6e6.7z: removeLink: does not exist (No such file or directory)]
but it do not raises errors without the -J 8
flag.
My 2cent is that one worker tries to get the archive, while another tries to get a file from the archive thus gets the archive too, causing conflict.
While getting files from a single archive is clearly not parallelizable, getting files with glob from multiple archives in parallel could make sense.
What steps will reproduce the problem?
create a dataset, add an archive, add archive content, push archive to a remote, drop all, try parallel get.
DataLad information
datalad 0.19.5
Additional context
No response
Have you had any success using DataLad before?
indeed ;D