What was reported:
Issue hashicorp/vault#22493 (closed, 21 comments): **Describe the bug**
When trying to migrate from GCS backend to some other backend (raft initially, but same issue happens with file), the migration process fails with `context canceled` after some time. When using the following configuration file :
```
storage_source "gcs" {
bucket = "test-vault-migation"
}
storage_destination "file" {
path = "data"
}
cluster_addr = "http://the loopback address:8201"
```
on a VM with enough CPU/RAM/disk, located in same region than bucket, the first migration run will fail after ~37.5k items synced (this is consistent over time / retry, they all...
What works:
If vault operator migrate dies with 'context canceled' partway through, it may be a single oversized entry: raft's default max_entry_size is 1MB and the migration swallows the real error. The reporter fixed it by finding and deleting the offending entry (the debug logs showed which key was copied last before the failure). Alternative: raise max_entry_size in the raft storage config. Run the migration with -log-level=debug and watch the last copied key to find the culprit.