In Tumbleweed with have roughly ~25.000 packages for 64-bit architecture. Do you know how many of those are actually install-able? From those who are not, do you know the reason behind? Do you know how many of those will become install-able if boo#123456 gets fixed? And from those which can actually be installed, do you know if there are any glitches at the post-installation scripts?
-Sure, we have openQA, but still, it tests only the packages inside the DVD and not the entire ecosystem.
- Sure, we have the OBS. So, everything that gets build should also be install-able. No?
-Sure, we have libsolv techniques that can answer this. But have you tested if the results reflect the real world?
There's only way to do verify what's really happening: one system per package. Yes, that is extreme, you would probably need 25.000 virtual machines. But ... hold on... what about using containers? Well, I have an idea! I have developed a project for fun, and I would be delighted to share it with you.
Egkatastasis (you can call it *egg*) is an open source system for testing openSUSE container images providing basic mechanisms for installation, log analysis, and metrics visualization of every package contained into the official repositories.
Egkatastasis tests production container workloads at scale using Docker and systemd-nspawn, combined with the best-of-breed ideas and practices from the community using Filebeat and Elastic Stack.
In Tumbleweed with have roughly ~25.000 packages for 64-bit architecture. Do you know how many of those are actually install-able? From those who are not, do you know the reason behind? Do you know how many of those will become install-able if boo#123456 gets fixed? And from those which can actually be installed, do you know if there are any glitches at the post-installation scripts?
-Sure, we have openQA, but still, it tests only the packages inside the DVD and not the entire ecosystem.
- Sure, we have the OBS. So, everything that gets build should also be install-able. No?
-Sure, we have libsolv techniques that can answer this. But have you tested if the results reflect the real world?
There's only way to do verify what's really happening: one system per package. Yes, that is extreme, you would probably need 25.000 virtual machines. But ... hold on... what about using containers? Well, I have an idea! I have developed a project for fun, and I would be delighted to share it with you.
Egkatastasis (you can call it *egg*) is an open source system for testing openSUSE container images providing basic mechanisms for installation, log analysis, and metrics visualization of every package contained into the official repositories.
Egkatastasis tests production container workloads at scale using Docker and systemd-nspawn, combined with the best-of-breed ideas and practices from the community using Filebeat and Elastic Stack.