Ignoring a repository while creating indexes

Background: for Reasons™ we have some big open source repos mirrored on our Phabricator instance. Everything works quite well, but some parts of Phabricator definitely suffer from worse user experience.

  1. DiffusionCommits database grows like mad with commits that are in no way interesting.
  2. Search with default settings is sometimes hard to use, since some search terms produce unholy amount of unrelated commits from those huge repos.
  3. After latest upgrade, the same happens with Identities, and our Identity pool is polluted with completely irrelevant entries.

I’m all ears for any kind of options that would alleviate these challenges.

I was thinking that, if not supported natively, maybe a very simple solution would be to blacklist those repositories for some of the mentioned indexing tasks just by hard-coding them in the source. How’s that sound, and anyone have any pointers on where I should poke the source to make something like this happen?

It kinda depends on your Reasons for mirroring locally… You can can Disable the repo, but then some of your Reasons might fail.
You can configure Branches and Permanent Refs to exclude all branches, or disable publishing completely; This will disable many features, but keep some (will still update the local copy, I guess?).