When a chunk server dies, its chunks become under-replicated. The master must automatically schedule re-replication to restore the target replication factor.

Re-replication algorithm:

1. Detect: master notices missing heartbeats from a server and marks it dead
2. Scan: identify all chunks that were on the dead server — they now have fewer replicas
3. Prioritize: chunks with RF=1 are critical (one more failure = data loss). Re-replicate them first.
4. Schedule: for each under-replicated chunk, pick a healthy server that does NOT already hold the chunk
5. Copy: instruct an existing replica to send the chunk data to the new server
6. Update: add the new server to the chunk's location list in the master's metadata

Request:  {"type": "check_replication", "msg_id": 1}
Response: {"type": "check_replication_ok", "in_reply_to": 1, "under_replicated": [
    {"chunk": "ch_001", "current_rf": 2, "target_rf": 3, "missing_on": ["cs3"]},
    {"chunk": "ch_005", "current_rf": 1, "target_rf": 3, "priority": "critical"}
]}

Request:  {"type": "replicate_chunk", "msg_id": 2, "chunk": "ch_005", "source": "cs1", "target": "cs4"}
Response: {"type": "replicate_chunk_ok", "in_reply_to": 2, "chunk": "ch_005", "new_rf": 2, "bytes_copied": 67108864}