[Request for Comments]
We recently got a zabbix alert about disk threshold usage on https://debuginfod.centos.org service. As a reminder (and for context), while all previous centos linux releases had (for efficiency reasons) a split between .src.rpm (going to https://vault.centos.org) and *debuginfo* rpm packages (consuming a lot of bandwidth/disk space) going to http://debuginfo.centos.org , the plan changed when Stream 8 (and so Stream 9 and 10 too) landed.
All packages were pushed out to mirror network, but not all variants (multiple ones due to Stream nature) were kept online.
We were asked to provide a similar service to what Fedora was already offering and so that's how https://debuginfod.centos.org/ came to life : for people willing to debug pkgs, they can request "on the fly" corresponding debuginfo files/packages (see website for simple instructions)
Now the question (and requests for comments) : We were still carrying on that server all the Stream 8 debuginfo packages, while technically not needed anymore (Stream 8 was removed from all mirrors when it went EOL in May 2024) ...
But based on analysis done with Frank Eigler (maintaining the service with us), we discovered that in fact a third of all requests for debuginfod service are still about stream 8 (vs Stream 9 and 10)
The question is so : can we just delete (imho, yes !) these old .el8. debuginfo packages (probably just analyzed in loop by AI scrapers these days) and so : - limit bandwidth needed for that machine - reclaim some storage space (instead of increasing in loop the backend disks)
Opinions, thoughts, comments ?