On 3/18/22 02:31, Stijn Vanden Brande (External) wrote:
The solution should be simply running `createrepo` after the new package is added in archive repository to have the correct repo data.
You stated the issue in your initial observation. The contents of archive.a.o are an automated ASF-wide rsync from downloads.a.o. There is no human to run `createrepo` on the archives. Nor would the addition of some sort of post-rsync script work for all Apache Software Foundation projects, as this would be error prone. It is a rolling archive of all ASF current release artifacts. That's it.
If users have a need to follow a specific version for their clusters, the recommendation has always been that they download the specific artifacts they need, verify sigs, etc. and toss them on their own web server, using the package repository tools they like for deb/rpm packages. It is very common for database clusters to have no direct internet access at all, and for nodes to have an internal-only web server in the infrastructure with their "golden" cluster package repository. Test test test some new version in dev or from the public repos, then upload those verified packages to the "golden" repo for production deployment. Done.
There is no simple way to do this on archive.a.o, so you'll have to do this yourself, which ends up being way more stable and predictable in the long run. It's a great question that comes up from time to time, it just hasn't been asked recently.
Warm regards, Michael