GEOS
GEOS copied to clipboard
Rollback azure pipeline, fine grained usage of travis VMs, travis pipeline cleaning
It appears that we were not using the VM sizes we were paying travis-ci
for.
- [x] Use the
large
machines for all thetravis
pipeline. Done in https://github.com/GEOSX/GEOSX/pull/2025 - [x] Rollback all the azure-pipeline actions:
git
history will allow to get back to the files if we want to go back to it. Done in https://github.com/GEOSX/GEOSX/pull/2025 - [x] We could add the
check_pr_is_assigned
in its own initial step such that we do not initiate jobs for nothing (100% waste). Done in https://github.com/GEOSX/GEOSX/pull/2053 - [x] https://config.travis-ci.com/explore offers an interactive way to test the
.travis.yaml
file, so we should endeavor to remove the warnings or informative messages (it should be a tiny action). Done in https://github.com/GEOSX/GEOSX/pull/2053 - [ ] Try to use specific sizes for each steps (tipically,
check
steps sould run onmedium
VMs, and maybe running the GPUs onx-large
?). I'm not sure that it's indeed possible, but https://config.travis-ci.com/explore seemed to consider it as an option, so let's try 🤞
Try to use specific sizes for each steps...
Yes, a github search shows size can be adjusted for each job launch.
I think we should rush https://github.com/GEOSX/GEOSX/pull/2025 for immediate relief of our build queues (tick box one). Maybe we can merge it as is and do a better cleaning after?
@rrsettgast I'm trying to finish the last tick box.
Whatever I may do in the config file, I stick to large
VMs.
Do you think we could ask the travis support to stop our specific VM choice so we may be able to better configure them?
According to our current .travis.yml
, it should be equivalent and seamless.
Actually, even if this does not allow a fine grain configuration, I also prefer relying on explicit documentation than on hidden backend configurations.
The use of larger VM's did not work.
The use of larger VM's did not work.
I know:
Whatever I may do in the config file, I stick to large VMs.
But is it because the travis
support imposes a specific VM size like they used to do when we couldn't chose the VM sizes ? Or is it because it's impossible?
Also, IIRC, I could not downsize a VM, which really surprises me and makes me think there's something else. Downsizing should really work.
Questions related to travis are not more relevant 😉