test(e2e): cover mixed-version rolling upgrades (#6975)

.github/workflows/e2e-upgrade.yml

@@ -49,8 +49,20 @@ env:
   UPGRADE_SOURCE_SHA256: 7c789386bf85278f865b8e0d359bf4edb84d5aa408cc3fa54a18c25ca74cd6e7
 
 jobs:
-  direct-upgrade:
-    name: Direct upgrade from rc.2
+  upgrade:
+    name: ${{ matrix.name }}
+    strategy:
+      fail-fast: false
+      matrix:
+        include:
+          - name: Direct upgrade from rc.2
+            cache_key: e2e-direct-upgrade
+            test: direct_upgrade_from_rc2_preserves_object_contracts
+            artifact: direct-upgrade
+          - name: Mixed-version rolling upgrade from rc.2
+            cache_key: e2e-mixed-version-upgrade
+            test: rolling_upgrade_from_rc2_preserves_mixed_version_contracts
+            artifact: mixed-version-upgrade
     runs-on: ubuntu-latest
     timeout-minutes: 60
     env:
@@ -64,7 +76,7 @@ jobs:
       - name: Setup Rust environment
         uses: ./.github/actions/setup
         with:
-          cache-shared-key: e2e-direct-upgrade
+          cache-shared-key: ${{ matrix.cache_key }}
           cache-save-if: ${{ github.ref == 'refs/heads/main' }}
           install-build-packaging-tools: "false"
 
@@ -89,17 +101,17 @@ jobs:
           cargo build --locked -p rustfs --bin rustfs
           : > target/debug/rustfs.features
 
-      - name: Run direct-upgrade compatibility test
+      - name: Run upgrade compatibility test
         run: |
           cargo test --locked -p e2e_test \
-            upgrade_compatibility_test::direct_upgrade_from_rc2_preserves_object_contracts \
+            "upgrade_compatibility_test::${{ matrix.test }}" \
             -- --ignored --exact --nocapture
 
       - name: Upload server logs
         if: always()
         uses: actions/upload-artifact@b7c566a772e6b6bfb58ed0dc250532a479d7789f # v6
         with:
-          name: direct-upgrade-server-logs-${{ github.run_number }}
+          name: ${{ matrix.artifact }}-server-logs-${{ github.run_number }}
           path: ${{ runner.temp }}/rustfs-upgrade-logs
           if-no-files-found: warn
           retention-days: 14

crates/e2e_test/README.md

@@ -169,7 +169,7 @@ the same profile for membership and execution with one nightly worker.
 | `s3s-e2e` black-box | `e2e-tests` + `e2e-tests-rio-v2` jobs | **Active** (external conformance tool) |
 | ILM / lifecycle (ignored) | `test-ilm-integration-serial` lane, `-j1` | **Active** (backlog#1148 ilm-1) |
 | KMS suite | `e2e-full` job, merge queue + main | **Active** |
-| Direct upgrade from pinned previous release | `e2e-upgrade.yml`, storage-sensitive PRs + release tags + weekly | **Active** |
+| Direct and mixed-version rolling upgrades from pinned previous release | `e2e-upgrade.yml`, storage-sensitive PRs + release tags + weekly | **Active** |
 | Cluster faults (`e2e-nightly` profile) | consolidated nightly workflow | **Active** (backlog#1149 ci-7) |
 | Protocols (FTPS/WebDAV/SFTP) | consolidated nightly workflow, serial | **Active** (backlog#1149 ci-7) |
 | Replication (fast subset) | `e2e-smoke` profile, `e2e-tests` job, every PR | **Active** (backlog#1147 repl-1) |

crates/e2e_test/src/common.rs

@@ -1469,31 +1469,18 @@ impl RustFSTestClusterEnvironment {
     ///   times out, or cluster service readiness times out.
     pub async fn start(&mut self) -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
         let binary_path = rustfs_binary_path();
+        self.start_with_binary(&binary_path).await
+    }
+
+    /// Start every cluster node with a specific RustFS binary.
+    ///
+    /// Upgrade compatibility tests use this to initialize a cluster with a
+    /// pinned previous release before replacing nodes with the workspace build.
+    pub async fn start_with_binary(&mut self, binary_path: &Path) -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
         let volumes_arg = self.build_volumes_arg();
 
-        for (i, node) in self.nodes.iter_mut().enumerate() {
-            info!("Starting cluster node {} on {}", i, node.address);
-
-            let mut command = Command::new(&binary_path);
-            command
-                .env("RUSTFS_VOLUMES", &volumes_arg)
-                .env("RUSTFS_ADDRESS", &node.address)
-                .env("RUSTFS_ACCESS_KEY", &self.access_key)
-                .env("RUSTFS_SECRET_KEY", &self.secret_key)
-                .env("RUSTFS_CONSOLE_ENABLE", "false")
-                .env("RUST_LOG", "rustfs=info,rustfs_notify=debug");
-
-            for (key, value) in &self.extra_env {
-                command.env(key, value);
-            }
-            for (key, value) in &self.node_extra_env[i] {
-                command.env(key, value);
-            }
-            capture_command_logs(&mut command, self.node_capture_log_paths[i].as_deref())?;
-
-            let process = command.current_dir(&node.data_dir).spawn()?;
-
-            node.process = Some(process);
+        for node_idx in 0..self.nodes.len() {
+            self.spawn_node(node_idx, binary_path, &volumes_arg)?;
         }
 
         for (i, node) in self.nodes.iter().enumerate() {
@@ -1509,20 +1496,46 @@ impl RustFSTestClusterEnvironment {
 
     /// Start one node process using the cluster's existing volume layout.
     pub async fn start_node(&mut self, node_idx: usize) -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
+        let binary_path = rustfs_binary_path();
+        self.start_node_from_binary(node_idx, &binary_path).await
+    }
+
+    /// Start one stopped cluster node with a specific RustFS binary while
+    /// preserving the cluster's volume layout and that node's data directory.
+    pub async fn start_node_from_binary(
+        &mut self,
+        node_idx: usize,
+        binary_path: &Path,
+    ) -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
+        let volumes_arg = self.build_volumes_arg();
+        self.spawn_node(node_idx, binary_path, &volumes_arg)?;
+
+        self.wait_for_node_ready(&self.nodes[node_idx].address, node_idx).await?;
+        self.wait_for_node_service_ready(node_idx).await?;
+        Ok(())
+    }
+
+    fn spawn_node(
+        &mut self,
+        node_idx: usize,
+        binary_path: &Path,
+        volumes_arg: &str,
+    ) -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
         self.ensure_node_index(node_idx)?;
         if self.nodes[node_idx].process.is_some() {
             return Err(format!("cluster node {node_idx} is already running").into());
         }
+        if !binary_path.is_file() {
+            return Err(format!("RustFS binary does not exist: {}", binary_path.display()).into());
+        }
 
-        let binary_path = rustfs_binary_path();
-        let volumes_arg = self.build_volumes_arg();
         let log_path = self.node_capture_log_paths[node_idx].clone();
         let node = &mut self.nodes[node_idx];
-        info!("Starting cluster node {} on {}", node_idx, node.address);
+        info!("Starting cluster node {} on {} with {}", node_idx, node.address, binary_path.display());
 
-        let mut command = Command::new(&binary_path);
+        let mut command = Command::new(binary_path);
         command
-            .env("RUSTFS_VOLUMES", &volumes_arg)
+            .env("RUSTFS_VOLUMES", volumes_arg)
             .env("RUSTFS_ADDRESS", &node.address)
             .env("RUSTFS_ACCESS_KEY", &self.access_key)
             .env("RUSTFS_SECRET_KEY", &self.secret_key)
@@ -1539,9 +1552,6 @@ impl RustFSTestClusterEnvironment {
 
         let process = command.current_dir(&node.data_dir).spawn()?;
         node.process = Some(process);
-
-        self.wait_for_node_ready(&self.nodes[node_idx].address, node_idx).await?;
-        self.wait_for_node_service_ready(node_idx).await?;
         Ok(())
     }
 

crates/e2e_test/src/upgrade_compatibility_test.rs

@@ -12,14 +12,15 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
-use crate::common::{RustFSTestEnvironment, init_logging};
+use crate::common::{RustFSTestClusterEnvironment, RustFSTestEnvironment, init_logging, rustfs_binary_path};
 use aws_sdk_s3::Client;
 use aws_sdk_s3::error::ProvideErrorMetadata;
 use aws_sdk_s3::primitives::ByteStream;
 use aws_sdk_s3::types::{
     BucketVersioningStatus, CompletedMultipartUpload, CompletedPart, ServerSideEncryption, VersioningConfiguration,
 };
-use std::path::PathBuf;
+use std::path::{Path, PathBuf};
+use tokio::task::JoinSet;
 
 type TestResult = Result<(), Box<dyn std::error::Error + Send + Sync>>;
 
@@ -28,6 +29,10 @@ const SSE_MASTER_KEY_ENV: &str = "RUSTFS_SSE_S3_MASTER_KEY";
 const SSE_MASTER_KEY: &str = "QkJCQkJCQkJCQkJCQkJCQkJCQkJCQkJCQkJCQkJCQkI=";
 const PLAIN_BUCKET: &str = "upgrade-plain-data";
 const VERSIONED_BUCKET: &str = "upgrade-versioned-data";
+const MIXED_BUCKET: &str = "upgrade-mixed-version-data";
+const MIXED_NODE_COUNT: usize = 4;
+const MULTIPART_WORKERS: usize = 16;
+const MULTIPART_UPLOADS_PER_WORKER: usize = 16;
 
 fn source_binary() -> Result<PathBuf, Box<dyn std::error::Error + Send + Sync>> {
     let path = std::env::var_os(SOURCE_BINARY_ENV)
@@ -103,6 +108,99 @@ async fn write_multipart(client: &Client, bucket: &str, key: &str, parts: &[Vec<
     Ok(())
 }
 
+fn configure_cluster_logs(cluster: &mut RustFSTestClusterEnvironment) -> TestResult {
+    let Some(log_dir) = std::env::var_os("RUSTFS_E2E_LOG_DIR") else {
+        return Ok(());
+    };
+    std::fs::create_dir_all(&log_dir)?;
+    for node_idx in 0..cluster.nodes.len() {
+        let path = Path::new(&log_dir).join(format!("mixed-upgrade-node-{node_idx}.log"));
+        cluster.set_node_capture_log_path(node_idx, path.to_string_lossy().into_owned())?;
+    }
+    Ok(())
+}
+
+async fn write_multipart_load(clients: &[Client], phase: &str) -> Result<Vec<String>, Box<dyn std::error::Error + Send + Sync>> {
+    let mut tasks = JoinSet::new();
+    for worker in 0..MULTIPART_WORKERS {
+        let client = clients[worker % clients.len()].clone();
+        let phase = phase.to_string();
+        tasks.spawn(async move {
+            let mut keys = Vec::with_capacity(MULTIPART_UPLOADS_PER_WORKER);
+            for upload in 0..MULTIPART_UPLOADS_PER_WORKER {
+                let key = format!("{phase}/multipart/{worker:02}/{upload:02}");
+                let part = vec![u8::try_from(worker)?; 64 * 1024];
+                write_multipart(&client, MIXED_BUCKET, &key, &[part]).await?;
+                keys.push(key);
+            }
+            Ok::<_, Box<dyn std::error::Error + Send + Sync>>(keys)
+        });
+    }
+
+    let mut keys = Vec::with_capacity(MULTIPART_WORKERS * MULTIPART_UPLOADS_PER_WORKER);
+    while let Some(result) = tasks.join_next().await {
+        keys.extend(result??);
+    }
+    Ok(keys)
+}
+
+async fn exercise_mixed_cluster(
+    cluster: &RustFSTestClusterEnvironment,
+    phase: &str,
+    current_node: usize,
+    previous_node: usize,
+) -> TestResult {
+    let clients = cluster.create_all_clients()?;
+    let current_client = &clients[current_node];
+    let previous_client = &clients[previous_node];
+
+    let current_key = format!("{phase}/written-by-current");
+    let current_body = format!("{phase}: current RustFS build").into_bytes();
+    current_client
+        .put_object()
+        .bucket(MIXED_BUCKET)
+        .key(&current_key)
+        .body(ByteStream::from(current_body.clone()))
+        .send()
+        .await?;
+    assert_eq!(read_object(previous_client, MIXED_BUCKET, &current_key, None).await?.1, current_body);
+
+    let previous_key = format!("{phase}/written-by-previous");
+    let previous_body = format!("{phase}: previous RustFS release").into_bytes();
+    previous_client
+        .put_object()
+        .bucket(MIXED_BUCKET)
+        .key(&previous_key)
+        .body(ByteStream::from(previous_body.clone()))
+        .send()
+        .await?;
+    assert_eq!(read_object(current_client, MIXED_BUCKET, &previous_key, None).await?.1, previous_body);
+
+    let multipart_keys = write_multipart_load(&clients, phase).await?;
+    let expected_count = multipart_keys.len() + 2;
+    for client in [current_client, previous_client] {
+        let listed = client
+            .list_objects_v2()
+            .bucket(MIXED_BUCKET)
+            .prefix(format!("{phase}/"))
+            .send()
+            .await?;
+        assert_eq!(
+            listed.contents().len(),
+            expected_count,
+            "both RustFS versions must stream the complete mixed-version listing"
+        );
+    }
+
+    let last_multipart_key = format!("{phase}/multipart/{:02}/{:02}", MULTIPART_WORKERS - 1, MULTIPART_UPLOADS_PER_WORKER - 1);
+    assert_eq!(
+        read_object(previous_client, MIXED_BUCKET, &last_multipart_key, None).await?.1,
+        vec![u8::try_from(MULTIPART_WORKERS - 1)?; 64 * 1024]
+    );
+
+    Ok(())
+}
+
 #[tokio::test]
 #[ignore = "requires a pinned previous RustFS release binary"]
 async fn direct_upgrade_from_rc2_preserves_object_contracts() -> TestResult {
@@ -252,3 +350,47 @@ async fn direct_upgrade_from_rc2_preserves_object_contracts() -> TestResult {
 
     Ok(())
 }
+
+#[tokio::test]
+#[ignore = "requires a pinned previous RustFS release binary"]
+async fn rolling_upgrade_from_rc2_preserves_mixed_version_contracts() -> TestResult {
+    init_logging();
+    let previous_binary = source_binary()?;
+    let current_binary = rustfs_binary_path();
+    let mut cluster = RustFSTestClusterEnvironment::new(MIXED_NODE_COUNT).await?;
+    cluster.set_env("RUST_LOG", "rustfs=warn,rustfs_notify=warn");
+    configure_cluster_logs(&mut cluster)?;
+    cluster.start_with_binary(&previous_binary).await?;
+    cluster.create_test_bucket(MIXED_BUCKET).await?;
+
+    cluster.stop_node(0)?;
+    cluster.start_node_from_binary(0, &current_binary).await?;
+    exercise_mixed_cluster(&cluster, "one-current-node", 0, 1).await?;
+
+    for node_idx in [1, 2] {
+        cluster.stop_node(node_idx)?;
+        cluster.start_node_from_binary(node_idx, &current_binary).await?;
+    }
+    exercise_mixed_cluster(&cluster, "one-previous-node", 0, 3).await?;
+
+    cluster.stop_node(3)?;
+    cluster.start_node_from_binary(3, &current_binary).await?;
+
+    for client in cluster.create_all_clients()? {
+        for phase in ["one-current-node", "one-previous-node"] {
+            let listed = client
+                .list_objects_v2()
+                .bucket(MIXED_BUCKET)
+                .prefix(format!("{phase}/"))
+                .send()
+                .await?;
+            assert_eq!(
+                listed.contents().len(),
+                MULTIPART_WORKERS * MULTIPART_UPLOADS_PER_WORKER + 2,
+                "the homogeneous current cluster must preserve every object from {phase}"
+            );
+        }
+    }
+
+    Ok(())
+}