Closed drosofff closed 4 years ago
I have some hopes that re-using miniconda role will fix many bugs, including the _upload issue... Yet have to understand why the playbook doesn't manage (anymore ?) the _upload nginx section in nginx.conf
Didn't look into this at all, but you don't need _upload
route in the nginx config or the nginx upload module anymore, Galaxy can now handle large uploads without that. When migrating away from that also make sure you don't have nginx_upload_store
and nginx_upload_path
set in the Galaxy config.
Hope that helps!
Thanks for the tip @mvdbeek
Over the 6 last months many changes have been introduced in gks and a number of bugs escapes to automated testing with Travis.
We should address:
[x] Upload is broken in the docker instance: work on the absence of _upload section in nginx.conf
[ ] A more appropriate job_conf.xml for quick use of the docker instance. This is also an issue with ansible
[ ] Slurm is not well working in the docker instance. Dig into this damned problem of supervisor in docker
[x] Go back to use of the miniconda role for conda, since its drop off is probably responsible for the small bug at tool installation with the install-tool role
[x] the dependency_resolvers_conf.xml file is not well installed, at least in 19.01