diff --git a/vlorentz/objstorage-replay-exclusion-file/README.txt b/vlorentz/objstorage-replay-exclusion-file/README.txt
new file mode 100644
index 0000000..aa2cc37
--- /dev/null
+++ b/vlorentz/objstorage-replay-exclusion-file/README.txt
@@ -0,0 +1,7 @@
+How to use:
+
+1. get unsorted_inventory.txt.gz
+2. sort it (--temporary-directory is needed if /tmp doesn't have >100GB free):
+   pv unsorted_inventory.txt.gz | pigz -d | LANG=C sort --temporary-directory=$HOME/tmp/ --parallel=12 --buffer-size=1G --compress-program=pigz | pv --wait --line-mode | pigz > sorted_inventory.txt.gz
+3. convert hex digests to binary:
+   pv sorted_inventory.txt.gz | pigz -d | python3 hash_txt_to_bytes.py > /srv/softwareheritage/cassandra-test-0/scratch/sorted_inventory.bin
diff --git a/vlorentz/objstorage-replay-exclusion-file/hash_txt_to_bytes.py b/vlorentz/objstorage-replay-exclusion-file/hash_txt_to_bytes.py
new file mode 100644
index 0000000..2f9367d
--- /dev/null
+++ b/vlorentz/objstorage-replay-exclusion-file/hash_txt_to_bytes.py
@@ -0,0 +1,19 @@
+import sys
+
+SHA1_CHARS = frozenset('0123456789abcdef')
+
+while True:
+    line = sys.stdin.readline()
+    if line == '\n':
+        break
+    assert line[-1] == '\n', repr(line)
+    if not set(line[0:-1]) <= SHA1_CHARS or len(line) != 41:
+        if len(line) != 3:
+            print('Rejecting: %r' % line, file=sys.stderr)
+        continue
+    try:
+        sys.stdout.buffer.write(bytes.fromhex(line[0:40]))
+    except Exception:
+        print(repr(line), file=sys.stderr)
+        raise
+