test(interop): add real MinIO read and migration parity tests (#4377)

test(interop): real-MinIO read + migration parity, Phase 1/2 (backlog#580)

Capture authentic on-disk fixtures from MinIO RELEASE.2025-07-23 (a bucket with
versioning, object-lock, lifecycle, tagging, quota, a public policy, SSE-S3
encryption, a webhook notification target, and a replication rule, plus inline /
versioned / multipart objects and a delete marker) and prove RustFS reads and
migrates them losslessly:

- filemeta parses_real_minio_object_xlmeta: small inline, two-object-version +
  delete marker, and multipart object xl.meta parse to the expected FileInfo.
- ecstore parses_real_minio_bucket_metadata_blob_without_loss: the MinIO
  .metadata.bin msgpack decodes via the PascalCase field names and
  parse_all_configs loads all ten config types present (policy, lifecycle incl.
  <ExpiryUpdatedAt>, object-lock, versioning, tagging, quota, notification,
  encryption/SSE-S3, replication incl. DeleteMarkerReplication /
  ExistingObjectReplication) without loss.
- ecstore reads_minio_inline_bucket_metadata_via_bitrot: MinIO inlines an object
  body as [HighwayHash256 32B][body]; RustFS's BitrotReader with HighwayHash256S
  verifies and yields the exact blob (the "inline_data 前缀不同" is that prefix).
- ecstore migrates_real_minio_bucket_metadata_end_to_end: on a throwaway 4-drive
  local ECStore, a real MinIO .metadata.bin seeded under .minio.sys is migrated
  into .rustfs.sys byte-identically for every config, exercising the Phase 2
  source adapter (MIGRATING_META_BUCKET = ".minio.sys") through the object layer.

All four run as ordinary crate tests (nextest CI). Phase 4 (MinIO re-reading a
RustFS drive) is documented as out of scope for one-way migration.

Refs rustfs/backlog#580

crates/ecstore/src/bucket/metadata.rs

@@ -1100,6 +1100,98 @@ where
 mod test {
     use super::*;
 
+    /// Decode a whitespace-tolerant hex fixture into bytes.
+    fn decode_hex(s: &str) -> Vec<u8> {
+        let s: String = s.chars().filter(|c| !c.is_whitespace()).collect();
+        (0..s.len())
+            .step_by(2)
+            .map(|i| u8::from_str_radix(&s[i..i + 2], 16).expect("valid hex fixture"))
+            .collect()
+    }
+
+    /// backlog#580: prove RustFS parses a real MinIO-written bucket `.metadata.bin`
+    /// blob without loss. The fixture is the raw `.metadata.bin` object body
+    /// (4-byte `format|version` header + msgpack) carved from the
+    /// `.minio.sys/buckets/interop/.metadata.bin` object written by MinIO
+    /// `RELEASE.2025-07-23` — see `tests/fixtures/minio/README.md`.
+    #[test]
+    fn parses_real_minio_bucket_metadata_blob_without_loss() {
+        let blob = decode_hex(include_str!("../../tests/fixtures/minio/bucket_metadata.blob.hex"));
+
+        // Same 4-byte format|version header (1|1) and msgpack layout as MinIO.
+        BucketMetadata::check_header(&blob).expect("valid .metadata.bin header");
+        let mut bm = BucketMetadata::unmarshal(&blob[4..]).expect("unmarshal MinIO bucket metadata");
+
+        // Raw config fields survive the msgpack decode (PascalCase MinIO field names).
+        assert_eq!(bm.name, "interop");
+        assert!(!bm.policy_config_json.is_empty(), "policy JSON present");
+        assert!(!bm.lifecycle_config_xml.is_empty(), "lifecycle XML present");
+        assert!(!bm.object_lock_config_xml.is_empty(), "object-lock XML present");
+        assert!(!bm.versioning_config_xml.is_empty(), "versioning XML present");
+        assert!(!bm.tagging_config_xml.is_empty(), "tagging XML present");
+        assert!(!bm.quota_config_json.is_empty(), "quota JSON present");
+
+        // Typed parse of each stored config must succeed. `parse_all_configs`
+        // logs+skips on error, so a None here means a real MinIO-compat parse gap.
+        bm.parse_all_configs().expect("parse_all_configs");
+        assert!(bm.policy_config.is_some(), "policy parsed");
+        assert!(bm.versioning_config.is_some(), "versioning parsed");
+        assert!(bm.object_lock_config.is_some(), "object-lock parsed");
+        assert!(bm.tagging_config.is_some(), "tagging parsed");
+        assert!(bm.quota_config.is_some(), "quota parsed");
+        assert!(
+            bm.lifecycle_config.is_some(),
+            "lifecycle parsed (MinIO writes an <ExpiryUpdatedAt> extension element)"
+        );
+        assert!(bm.notification_config.is_some(), "notification parsed");
+        assert!(bm.sse_config.is_some(), "encryption (SSE) parsed");
+        assert!(bm.replication_config.is_some(), "replication parsed");
+
+        // Object lock is expressed through the parsed config, not the legacy
+        // `LockEnabled` flag (MinIO leaves that false for config-based locks).
+        assert!(bm.object_locking(), "object lock active via parsed config");
+    }
+
+    /// backlog#580: KNOWN GAP (weisd 2026-03-06 "inline_data 前缀不同"). RustFS's
+    /// inline-data extraction does not yet recover the object body from a
+    /// MinIO-written bucket-metadata object: `into_fileinfo(read_data=true).data`
+    /// returns bytes that are not the `.metadata.bin` blob (no `format|version`
+    /// header). Kept as an ignored, documented reproduction until the MinIO
+    /// inline-data framing is handled on the read path.
+    /// backlog#580: prove RustFS reads a MinIO-written **inlined** bucket-metadata
+    /// object end-to-end. MinIO stores inline data as `[bitrot hash][object body]`
+    /// (the "`inline_data` 前缀不同" that weisd flagged on 2026-03-06 is that
+    /// bitrot prefix, not a format incompatibility). Running the raw inline shard
+    /// through RustFS's `BitrotReader` with the default `HighwayHash256S` must
+    /// verify the checksum and yield the exact `.metadata.bin` blob.
+    #[tokio::test]
+    async fn reads_minio_inline_bucket_metadata_via_bitrot() {
+        use crate::erasure::coding::BitrotReader;
+        use rustfs_utils::HashAlgorithm;
+
+        let xlmeta = decode_hex(include_str!("../../tests/fixtures/minio/bucket_metadata_full.xlmeta.hex"));
+        let fm = rustfs_filemeta::FileMeta::load(&xlmeta).expect("parse MinIO xl.meta");
+        let fi = fm
+            .into_fileinfo("interop", ".metadata.bin", "", true, false, false)
+            .expect("into_fileinfo");
+        // The raw inline shard is `[HighwayHash256 (32B)][object body]`.
+        let inline = fi.data.expect("inline shard present");
+        let algo = HashAlgorithm::HighwayHash256S;
+        let body_len = inline.len() - algo.size();
+
+        let mut reader = BitrotReader::new(std::io::Cursor::new(inline.to_vec()), body_len, algo, false);
+        let mut body = vec![0u8; body_len];
+        let read = reader.read(&mut body).await.expect("bitrot verify + read MinIO inline shard");
+        assert_eq!(read, body_len);
+
+        // The verified body is exactly the `.metadata.bin` blob, and it parses.
+        BucketMetadata::check_header(&body).expect("recovered body is a valid .metadata.bin");
+        let mut bm = BucketMetadata::unmarshal(&body[4..]).expect("unmarshal recovered blob");
+        assert_eq!(bm.name, "interop");
+        bm.parse_all_configs().expect("parse recovered configs");
+        assert!(bm.lifecycle_config.is_some());
+    }
+
     #[tokio::test]
     async fn marshal_msg() {
         // write_time(OffsetDateTime::UNIX_EPOCH).unwrap();

crates/ecstore/src/bucket/migration.rs

@@ -517,4 +517,122 @@ mod tests {
         assert_eq!(decoded.id, 123);
         assert_eq!(decoded.targets_map["arn:replication::1:dest"].resync_id, "reset-1");
     }
+
+    fn decode_hex(s: &str) -> Vec<u8> {
+        let s: String = s.chars().filter(|c| !c.is_whitespace()).collect();
+        (0..s.len())
+            .step_by(2)
+            .map(|i| u8::from_str_radix(&s[i..i + 2], 16).expect("valid hex fixture"))
+            .collect()
+    }
+
+    /// backlog#580 Phase 2/4: end-to-end proof that `try_migrate_bucket_metadata`
+    /// pulls a real MinIO-written bucket-metadata blob from a `.minio.sys` layout
+    /// into `.rustfs.sys`, and that the migrated blob loads every config. Uses a
+    /// throwaway 4-drive local ECStore. Fixture: `tests/fixtures/minio/README.md`.
+    #[tokio::test]
+    async fn migrates_real_minio_bucket_metadata_end_to_end() {
+        use crate::bucket::metadata::{BUCKET_METADATA_FILE, BucketMetadata};
+        use crate::config::com::read_config;
+        use crate::disk::endpoint::Endpoint;
+        use crate::disk::{BUCKET_META_PREFIX, MIGRATING_META_BUCKET, RUSTFS_META_BUCKET};
+        use crate::layout::endpoints::{EndpointServerPools, Endpoints, PoolEndpoints};
+        use crate::object_api::{ObjectOptions, PutObjReader};
+        use crate::storage_api_contracts::bucket::{BucketOperations, BucketOptions, MakeBucketOptions};
+        use crate::storage_api_contracts::object::{ObjectIO, ObjectOperations};
+        use crate::store::{ECStore, init_local_disks};
+        use rustfs_utils::path::SLASH_SEPARATOR;
+        use tokio::fs;
+        use tokio_util::sync::CancellationToken;
+        use uuid::Uuid;
+
+        let blob = decode_hex(include_str!("../../tests/fixtures/minio/bucket_metadata.blob.hex"));
+
+        // --- Stand up a throwaway 4-drive local ECStore. ---
+        let base = std::path::PathBuf::from(format!("/tmp/rustfs_minio_migrate_test_{}", Uuid::new_v4()));
+        let disk_paths: Vec<_> = (1..=4).map(|i| base.join(format!("disk{i}"))).collect();
+        for p in &disk_paths {
+            fs::create_dir_all(p).await.unwrap();
+        }
+        let mut endpoints = Vec::new();
+        for (i, p) in disk_paths.iter().enumerate() {
+            let mut ep = Endpoint::try_from(p.to_str().unwrap()).unwrap();
+            ep.set_pool_index(0);
+            ep.set_set_index(0);
+            ep.set_disk_index(i);
+            endpoints.push(ep);
+        }
+        let endpoint_pools = EndpointServerPools(vec![PoolEndpoints {
+            legacy: false,
+            set_count: 1,
+            drives_per_set: 4,
+            endpoints: Endpoints::from(endpoints),
+            cmd_line: "minio-migrate-test".to_string(),
+            platform: format!("OS: {} | Arch: {}", std::env::consts::OS, std::env::consts::ARCH),
+        }]);
+        init_local_disks(endpoint_pools.clone()).await.unwrap();
+        let ecstore = ECStore::new("127.0.0.1:0".parse().unwrap(), endpoint_pools, CancellationToken::new())
+            .await
+            .unwrap();
+        let existing: Vec<String> = ecstore
+            .list_bucket(&BucketOptions {
+                no_metadata: true,
+                ..Default::default()
+            })
+            .await
+            .unwrap()
+            .into_iter()
+            .map(|b| b.name)
+            .collect();
+        crate::bucket::metadata_sys::init_bucket_metadata_sys(ecstore.clone(), existing).await;
+
+        let meta_path = format!("{BUCKET_META_PREFIX}{SLASH_SEPARATOR}interop{SLASH_SEPARATOR}{BUCKET_METADATA_FILE}");
+        let put_opts = ObjectOptions::default();
+
+        // --- Arrange the pre-migration state a MinIO import starts from: the ---
+        // bucket exists and its config lives under `.minio.sys`, while `.rustfs.sys`
+        // has no metadata for it yet.
+        ecstore.make_bucket("interop", &MakeBucketOptions::default()).await.unwrap();
+        let _ = ecstore
+            .delete_object(RUSTFS_META_BUCKET, &meta_path, ObjectOptions::default())
+            .await;
+        // MinIO leaves its `.minio.sys` meta volume on every drive; recreate it so
+        // the source object can be seeded through the object layer.
+        for p in &disk_paths {
+            fs::create_dir_all(p.join(MIGRATING_META_BUCKET)).await.ok();
+        }
+        let mut src = PutObjReader::from_vec(blob.clone());
+        ecstore
+            .put_object(MIGRATING_META_BUCKET, &meta_path, &mut src, &put_opts)
+            .await
+            .expect("seed .minio.sys bucket metadata");
+
+        // --- Run the real startup migration. ---
+        super::try_migrate_bucket_metadata(ecstore.clone()).await;
+
+        // --- The migrated `.rustfs.sys` blob must carry every MinIO config, ---
+        // byte-identical to the source (typed XML/JSON parsing of these fields is
+        // covered by the bucket-metadata parse-parity test).
+        let migrated = read_config(ecstore.clone(), &meta_path)
+            .await
+            .expect("read migrated bucket metadata");
+        BucketMetadata::check_header(&migrated).expect("migrated blob has a valid header");
+        let bm = BucketMetadata::unmarshal(&migrated[4..]).expect("unmarshal migrated bucket metadata");
+        let src = BucketMetadata::unmarshal(&blob[4..]).expect("unmarshal source bucket metadata");
+
+        assert_eq!(bm.name, "interop");
+        assert_eq!(bm.policy_config_json, src.policy_config_json, "policy migrated intact");
+        assert_eq!(bm.lifecycle_config_xml, src.lifecycle_config_xml, "lifecycle migrated intact");
+        assert_eq!(bm.object_lock_config_xml, src.object_lock_config_xml, "object-lock migrated intact");
+        assert_eq!(bm.versioning_config_xml, src.versioning_config_xml, "versioning migrated intact");
+        assert_eq!(bm.tagging_config_xml, src.tagging_config_xml, "tagging migrated intact");
+        assert_eq!(bm.quota_config_json, src.quota_config_json, "quota migrated intact");
+        assert_eq!(bm.notification_config_xml, src.notification_config_xml, "notification migrated intact");
+        assert_eq!(bm.encryption_config_xml, src.encryption_config_xml, "encryption migrated intact");
+        assert_eq!(bm.replication_config_xml, src.replication_config_xml, "replication migrated intact");
+        assert!(!bm.lifecycle_config_xml.is_empty(), "lifecycle present");
+        assert!(!bm.replication_config_xml.is_empty(), "replication present");
+
+        fs::remove_dir_all(&base).await.ok();
+    }
 }