3 ms·

Plenty of alternative approaches that can get you there. Here’s one: * distributed file system like gluster or ceph for repos * clustered db (eg replicated

• elevation 6y ago
  Load balancing redundant gitea over clustered postgres and clustered fs provides a resilient read-only stack. The trouble comes when the system receives two simultaneous pushes to the same branch. When ceph goes to merge them, which one wins? There has to be a distributed write mutex. Perhaps this mutex could be acquired in a pre-commit hook on the gitea nodes, but it's absolutely necessary (in addition to the other clustered services) to prevent silent data loss or corruption.
• rovr138 6y ago
  post-receive hook pushing everything to other instances. How the push happens could be git remotes, rsync, or whatever else you want.