These packages contains one or more xz-compressed tarball of "kernel runner environment" Docker volume archives for several Linux distributions optimized for running a small coordinator daemon ("kernel runner") inside arbitrary Docker images on Backend.AI clusters.
For instance, backend.ai-krunner-ubuntu package contains environment images for Ubuntu 16.04 and 18.04. Each tar.xz archives takes about 20-30 MiB, and we are going to add a few more in the near future (e.g., Ubuntu 20.04 / CentOS 8 / Alpine 3.11).
Of course, we won't accumulate the archives infinitely but keep only latest 2 or 3 versions of each Linux distributions. We also have been trying to minimize the size using Docker's multi-stage builds, xz compression, and bundling only the isolated Python and package directory (each volume archives are mounted as /opt/backend.ai in user containers).
Still, adding more packages and distribution supports to the environment volume archives may touch the package size limit of 60 MiB, and I believe that 120 MiB would be good for now. I am writing this request since our newly built backend.ai-krunner-alpine package just hit 60.1 MiB and I got upload failure.
Project
Size of release
120 MiB for each of the three packages
Which indexes
PyPI only.
Reasons for the request
These packages contains one or more xz-compressed tarball of "kernel runner environment" Docker volume archives for several Linux distributions optimized for running a small coordinator daemon ("kernel runner") inside arbitrary Docker images on Backend.AI clusters.
For instance,
backend.ai-krunner-ubuntu
package contains environment images for Ubuntu 16.04 and 18.04. Each tar.xz archives takes about 20-30 MiB, and we are going to add a few more in the near future (e.g., Ubuntu 20.04 / CentOS 8 / Alpine 3.11).Of course, we won't accumulate the archives infinitely but keep only latest 2 or 3 versions of each Linux distributions. We also have been trying to minimize the size using Docker's multi-stage builds, xz compression, and bundling only the isolated Python and package directory (each volume archives are mounted as
/opt/backend.ai
in user containers).Still, adding more packages and distribution supports to the environment volume archives may touch the package size limit of 60 MiB, and I believe that 120 MiB would be good for now. I am writing this request since our newly built
backend.ai-krunner-alpine
package just hit 60.1 MiB and I got upload failure.