Geo - Retry checksum calculation for failures on the primary node
We should retry to calculate the checksum for repositories where the calculation failed. Most of the failures look like transient Gitaly errors:
- 13:CalculateChecksum: gitCommand: process spawn timed out after 10s
- 14:Connect Failed
- 14:OS Error
- 14:Socket closed
- 4:Deadline Exceeded
Projects with high activity have a chance to heal themselves because we reset the checksum on every push, but projects with no/less activity need a manual interaction.
Edited by Douglas Barbosa Alexandre