I am running a Storj node on Windows. My node is actively running the migration process to Hashstore, and it is currently moving the remaining data from the old blobs directory.
However, the piecemigrate:chore process is stuck in an infinite loop, constantly throwing the following error every few seconds:
2026-06-11T20:40:22+02:00 INFO piecemigrate:chore couldn’t migrate {“error”: “getting the piece header: pieces error: PieceHeader framing field claims impossible size of 27576 bytes”, “sat”: “121RTSDpyNZVcEU84Ticf2L1ntiuUimbWgfATz21tuvgk3vzoA6”, “id”: “DH5B56FDO3FCIABMRKCPMFPGT6AJ5Q52NELUNYP7TWKZGXB45B2A”}
What I have checked so far:
I searched for the specific piece ID DH5B56... using a master file table search across the entire drive and within the storage/blobs/ directory, but the file cannot be found.
I checked the specific satellite folder mentioned in the log (121RTSD...) inside storage/blobs/. The old blobs subfolder for this specific satellite does not exist anymore—only the new hashstore folder exists for it. This indicates that the physical migration for this particular satellite already finished.
I even tried creating a dummy 0 KB file matching that piece ID structure inside a manual blobs/121RTSD.../dh/ path to see if the node would skip it, but the error persists with the exact same message.
It seems the node’s internal database or Hashstore index still holds a reference to this corrupted piece ID for a satellite that has already completed its folder migration, causing an unresolvable loop.
How can I force the node to drop or skip this ghost/corrupted piece so the migration chore can continue with the rest of my data?
Look like the problem are new, you can try this, clone storj source code, then you can modify some of the code before that line and compile a new storagenode with this command:
# go to storj source code
go build -o bin/storagenode cmd/storagenode # the binary is in ./bin/storagenode
Then rerun piecemigrate:chore to see and keep at it :).
I tried searching by name using the paths you provided in the link, and I saw there was a file inside the dh folder, so I tried deleting it from that path. It’s not giving an error. That looks promising.
What did you try specifically? What error did you receive specifically?
Nobody has a crystal ball, and if filesystem is not healthy, let alone to the point of its own repair toll bailing out, there is absolutely no reason to try to look any anything else.
Sorry, I didn’t specify earlier, but when I ran chkdsk, it repaired some files on the disk. However, the “piecemigrate:chore” migration was still causing problems; it kept trying to perform the migration in a loop.
After deleting the problematic file from the specified path, the issue has been resolved. The solution is in the link you provided, as it shows the correct paths; the log had suggested that the file was in a different folder.