Hi! 👋
While me and @yakutovicha were testing a new release of AiiDAlab docker stack (https://github.com/aiidalab/aiidalab-docker-stack), I noticed that installing the aiidalab-flexpart app unfortunately breaks the container. Specifically, the installation fails, since the verdi daemon restart command fails, which is because the aiida-flexpart has a pin on a the psycopg2<2.9. This in turn causes a downgrade of psycopg2-binary for v2.9.9 to 2.8.6. And for some reason, this version breaks verdi. In order to restore the container functionality, one needs to manually uninstall the older version by running
pip uninstall psycopg2-binary
Sorry if the above was super confusing. tl;dr: You can still use the app, but you need to have an older image with older aiida-core version. I don't know what is the last version that works, you'll have to test is, e.g. by downloading the following image docker.io/aiidalab/full-stack/aiida-2.4.3.
You can find all the existing image tags here: https://hub.docker.com/r/aiidalab/full-stack/tags?name=aiida
Ultimately, the issue needs to be solved in aiida-flexpart by relaxing the psycopg2-binary pin.
Hi! 👋
While me and @yakutovicha were testing a new release of AiiDAlab docker stack (https://github.com/aiidalab/aiidalab-docker-stack), I noticed that installing the
aiidalab-flexpartapp unfortunately breaks the container. Specifically, the installation fails, since theverdi daemon restartcommand fails, which is because theaiida-flexparthas a pin on a thepsycopg2<2.9. This in turn causes a downgrade ofpsycopg2-binaryfor v2.9.9 to 2.8.6. And for some reason, this version breaks verdi. In order to restore the container functionality, one needs to manually uninstall the older version by runningSorry if the above was super confusing. tl;dr: You can still use the app, but you need to have an older image with older aiida-core version. I don't know what is the last version that works, you'll have to test is, e.g. by downloading the following image
docker.io/aiidalab/full-stack/aiida-2.4.3.You can find all the existing image tags here: https://hub.docker.com/r/aiidalab/full-stack/tags?name=aiida
Ultimately, the issue needs to be solved in
aiida-flexpartby relaxing thepsycopg2-binarypin.