From Mageia wiki
Restarting builds on the build system
If a build fails on the build system due to a temporary issue (e.g. full disc, bad permissions on a file) it can be restarted once that issue is fixed. This saves time, space and spec file churn, especially for packages that take a long time to build.
While some architectures are still building
If the build is still ongoing for other architectures, run:
rm -f ~schedbot/uploads/done/cauldron/core/release/{buildid}_{arch}.fail
where {buildid} is the build ID and {arch} is the architecture to restart. ulri will see that there are no results (success or fail) for this architecture and restart the build.
After all architectures have already finished
If the build has already completed for all architectures, an architecture rebuild must be manually restarted by running:
su - schedbot ./secondary_submit /distrib/bootstrap/distrib/cauldron/SRPMS/core/release/{foo}.src.rpm cauldron {arch} core/release {user who should get the email in case of failure}