/bytes redundancy headers ignored

[ldeffenb]

Context

2.1.0-rc2 in sepolia testnet

Summary

When doing a single-chunk /bytes retrieval, it seems that the redundancy headers are being ignored and multiple chunks at different references are being fetched and cancelled anyway.

Expected behavior

If I specify Swarm-Redundancy-Strategy: 0 and Swarm-Redundancy-Fallback-Mode: false, I would expect only the referenced chunk to be retrieved.

Actual behavior

For every request, multiple chunks are being fetched and cancelled per the logs below. I have confirmed that the headers are visible in the api based on the first log in my hacked version. https://github.com/ldeffenb/bee/blob/2ebe25af0c8238a37f9526584f31b0bd00b6296a/pkg/api/bzz.go#L538

This is without the redundancy headers:
curl http://192.168.10.38:1633/bytes/6f6aae5e8bda728b1d50594204376c53d73453b8d195ab7b602a4df8ccaa2ecd --output temp.tmp
"time"="2024-05-24 14:36:23.851095" "level"="debug" "logger"="node/api" "msg"="downloadHandler" "reference"="6f6aae5e8bda728b1d50594204376c53d73453b8d195ab7b602a4df8ccaa2ecd" "Strategy"=null "Fallback"=null
"time"="2024-05-24 14:36:23.856788" "level"="debug" "logger"="node/retrieval" "msg"="retrieval failed" "chunk_address"="ab2cfc3671935481246dd99f7e3829c0ef678d69171e6816ef50b94b5a16d011" "error"="context canceled"
"time"="2024-05-24 14:36:23.856969" "level"="debug" "logger"="node/retrieval" "msg"="retrieval failed" "chunk_address"="7bb8257947db81bd1722918a7bbd3c1ebd8483cc561a0e2c7521c627a11c7ecc" "error"="context canceled"

The same happens with the headers supplied (note the downloadHandler sees the values):
curl -H "Swarm-Redundancy-Strategy: 0" -H "Swarm-Redundancy-Fallback-Mode: false" http://192.168.10.38:1633/bytes/6f6aae5e8bda728b1d50594204376c53d73453b8d195ab7b602a4df8ccaa2ecd --output temp.tmp
"time"="2024-05-24 14:39:39.053943" "level"="debug" "logger"="node/api" "msg"="downloadHandler" "reference"="6f6aae5e8bda728b1d50594204376c53d73453b8d195ab7b602a4df8ccaa2ecd" "Strategy"=0 "Fallback"=false
"time"="2024-05-24 14:39:39.054327" "level"="debug" "logger"="node/retrieval" "msg"="retrieval failed" "chunk_address"="ab2cfc3671935481246dd99f7e3829c0ef678d69171e6816ef50b94b5a16d011" "error"="context canceled"
"time"="2024-05-24 14:39:39.054532" "level"="debug" "logger"="node/retrieval" "msg"="retrieval failed" "chunk_address"="7bb8257947db81bd1722918a7bbd3c1ebd8483cc561a0e2c7521c627a11c7ecc" "error"="context canceled"

Steps to reproduce

Find a single chunk reference (like 6f6aae5e8bda728b1d50594204376c53d73453b8d195ab7b602a4df8ccaa2ecd in sepolia testnet) and do a /bytes retrieval on it. Watch your --verbosity 5 (Debug) logs for other chunk_address references failing with context canceled.

Possible solution

Not sure yet. I haven't figured out how to enable the redundancy logger and haven't had time to provide a known-enabled logger to see what's going on in there. https://github.com/ethersphere/bee/blob/8645aca8600408f9d9066460b756bc1c7e8b5c81/pkg/file/redundancy/getter/getter.go#L68