From f9cf4ff2bf1d3bde60849b3c69ef81ad5517e2ff Mon Sep 17 00:00:00 2001
From: cxymds <cxymds@gmail.com>
Date: Thu, 2 Jul 2026 22:36:48 +0800
Subject: [PATCH] fix(ecstore): surface multipart cleanup delete errors (#4198)

---
 .../src/bucket/lifecycle/bucket_lifecycle_ops.rs     |  3 +++
 crates/ecstore/src/set_disk/list.rs                  | 12 ++++++++++++
 2 files changed, 15 insertions(+)

diff --git a/crates/ecstore/src/bucket/lifecycle/bucket_lifecycle_ops.rs b/crates/ecstore/src/bucket/lifecycle/bucket_lifecycle_ops.rs
index d0d69b75c..e05421858 100644
--- a/crates/ecstore/src/bucket/lifecycle/bucket_lifecycle_ops.rs
+++ b/crates/ecstore/src/bucket/lifecycle/bucket_lifecycle_ops.rs
@@ -1658,6 +1658,9 @@ async fn cleanup_stale_multipart_uploads_in_set(set: &Arc<SetDisks>, now: Offset
     let mut deleted = 0usize;
     let mut candidates = HashMap::new();
 
+    // Discovery is intentionally local-owner based: each server lists the disks
+    // it owns locally. Once a stale upload path is found, delete_all fans out
+    // idempotently across the set to remove matching shards on every disk.
     for disk in set.get_local_disks().await.into_iter().flatten() {
         if !disk.is_online().await {
             continue;
diff --git a/crates/ecstore/src/set_disk/list.rs b/crates/ecstore/src/set_disk/list.rs
index d729eafd6..2fba03a7b 100644
--- a/crates/ecstore/src/set_disk/list.rs
+++ b/crates/ecstore/src/set_disk/list.rs
@@ -54,6 +54,18 @@ impl SetDisks {
             }
         }
 
+        let failed = errors.iter().filter(|err| err.is_some()).count();
+        if failed > 0 {
+            debug!(
+                bucket = %bucket,
+                prefix = %prefix,
+                failed,
+                total = errors.len(),
+                errors = ?errors,
+                "delete_all completed with disk errors"
+            );
+        }
+
         Ok(())
     }
 }