After months of tinkering, a server accumulates manual packages and third-party repos. Audit, purge and upgrade to get back to a clean, reproducible base.
A server running for a long time carries layers of changes no one remembers. Packages installed by hand, repos added for a single dependency, leftovers from tests. Flattening it is often safer than continuing to stack.
Audit before you purge
Start with the inventory. List the manually installed packages, distinct from those pulled by the distribution. List the active third-party repos, the ones that do not come from official mirrors. Every entry must justify itself or go.
Many manual packages date from a one-off need long since resolved. A forgotten third-party repo can pull a package that overrides a system version and breaks an update.
Purge, upgrade, clean up
Remove the useless third-party repos and the packages they served. Let the distribution take back control of versions. Apply the kernel and base package upgrade to get back onto a supported line.
Then clean up the leftovers: orphaned configs, files left by uninstalled packages, caches. A clean base is documented: what is installed, why, and where it comes from.
The principle: an operable server is a reproducible server. If the current state cannot be re-explained, bring it back to a known base before you have to under pressure.