No use of checksum comparison after replication is done
Once the replication is done (update of the repository at the gitaly storage) praefect runs CalculateChecksum for source and target gitaly nodes.
The result of the execution compared and if there is no match the error message would be written into the log (example):
{
"msg": "checksums do not match",
"hostname": "praefect-01-stor-gprd",
"shard": "default",
"level": "error",
"primary": {
"storage_name": "file-praefect-02",
"relative_path": "@hashed/fa/53/fa539965395b8382145f8370b34eab249cf610d2d6f2943c95b9b9d08a63d4a3.git"
},
"replica": {
"relative_path": "@hashed/fa/53/fa539965395b8382145f8370b34eab249cf610d2d6f2943c95b9b9d08a63d4a3.git",
"storage_name": "file-praefect-01"
},
"environment": "gprd",
"pid": 32412,
"time": "2020-09-03T20:58:57.840Z",
"tier": "sv",
"type": "praefect",
"fqdn": "praefect-01-stor-gprd.c.gitlab-production.internal",
"tag": "praefect",
"stage": "main"
}This happens pretty often:
Start of the replication: 
Errors about not matching checksums: 
As a first step we could properly propagate correlation_id to all logging entries to get more info about what is going on as it is not used in all places where we log something.
It is not clear if we should take any actions in regard of different checksums for the repository.