Add git size metrics to usage ping
Git can handle very large repos, but most aren't anywhere near the size of Linux or Windows. We should add metrics to the usage ping to understand the distribution of Git repo sizes. This will help us understand the type of optimizations and improvements will most help customers.
Proposal
At some frequency, we should calculate statistics (maybe using git-sizer) about each repository and store these in the database.
- size on disk
- number of commits
- number of trees
- number of blobs
- number of refs
Using these numbers we can calculate average
, std-dev
, max
for each metric across the instance and add this to the usage ping.