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
{{ message }}
This repository has been archived by the owner on Feb 3, 2021. It is now read-only.
this happened because zipp had released a new version 2.2.0 on 8th Feb which was not compatible
We hot-fixed it by preinstalling zipp==1.1.0 in setup_hosts.sh
# ensure these packages are compatibile before upgrading
python3 -m pip install zipp=="1.1.0"
python3 -m pip install pip=="18.0" pipenv=="2018.7.1"
But then again on 12th Feb a new version of virtualenv was released and we faced the same issue again - same issue caused by the release of a different package (virtualenv released a new version in Feb 11)
You are correct, locking the transitive dependencies is the only (short-term) solution for this issue. The hotfix you put forward will be merged and released.
Context
as part of spark cluster launch on batch, Aztk runs setup_host.sh script and as part of it some python packages are installed
in particular, this line where pipenv is installed - https://github.com/Azure/aztk/blob/master/aztk/node_scripts/setup_host.sh#L153
This step has failed for us twice over the past 2 days
Issue
pipenv=="2018.7.1" has open bounded transitive dependencies and new version releases to these packages causes the pipenv install to be very unstable
for example: On 11th Feb we faced this error:
this happened because zipp had released a new version 2.2.0 on 8th Feb which was not compatible
We hot-fixed it by preinstalling zipp==1.1.0 in setup_hosts.sh
But then again on 12th Feb a new version of virtualenv was released and we faced the same issue again - same issue caused by the release of a different package (virtualenv released a new version in Feb 11)
we have again hot-fixed it in a similar way
Solution
At the moment it looks like the only solution is to lock the transitive dependency version for all dependent packages as well
Please recommend if there is a better way of mitigating this issue
The text was updated successfully, but these errors were encountered: