Skip to content

SeaweedFS is a distributed storage system for object storage (S3), file systems, Iceberg table buckets, and Lance vector buckets, designed to handle billions of files with O(1) disk access and effortless horizontal scaling.

topology: clamp the deleted-vs-total subtractions in volume stats (#10633)

Chris Lu08f0ba55645f
diff --git a/weed/shell/command_fs_merge_volumes.go b/weed/shell/command_fs_merge_volumes.go
index 08288c0c7a..4bc256a35c 100644
--- a/weed/shell/command_fs_merge_volumes.go
+++ b/weed/shell/command_fs_merge_volumes.go
@@ -465,7 +465,12 @@ func (c *commandFsMergeVolumes) getVolumeSizeBasedOnPlan(plan map[needle.VolumeI
return size
}
+// getVolumeSize is the volume's live data size, clamped since
+// DeletedByteCount can transiently exceed Size.
func (c *commandFsMergeVolumes) getVolumeSize(volume *master_pb.VolumeInformationMessage) uint64 {
+ if volume.Size < volume.DeletedByteCount {
+ return 0
+ }
return volume.Size - volume.DeletedByteCount
}
diff --git a/weed/shell/command_fs_merge_volumes_test.go b/weed/shell/command_fs_merge_volumes_test.go
index e368f13439..21a71a377a 100644
--- a/weed/shell/command_fs_merge_volumes_test.go
+++ b/weed/shell/command_fs_merge_volumes_test.go
@@ -123,3 +123,18 @@ func TestCreateMergePlan_DirectedRejectsIneligible(t *testing.T) {
})
}
}
+
+// A volume reporting more deleted bytes than it holds must read as empty.
+func TestGetVolumeSize_ClampsDeletedOverSize(t *testing.T) {
+ c := newMergeCmd(250000)
+
+ overDeleted := &master_pb.VolumeInformationMessage{Id: 1, Size: 1000, DeletedByteCount: 4000}
+ if got := c.getVolumeSize(overDeleted); got != 0 {
+ t.Errorf("expected 0 for a volume with more deleted than stored, got %d", got)
+ }
+
+ healthy := &master_pb.VolumeInformationMessage{Id: 2, Size: 3000, DeletedByteCount: 1000}
+ if got := c.getVolumeSize(healthy); got != 2000 {
+ t.Errorf("expected 2000, got %d", got)
+ }
+}
diff --git a/weed/topology/topology_stats_test.go b/weed/topology/topology_stats_test.go
index e4bee9bd6b..7b2c186ce3 100644
--- a/weed/topology/topology_stats_test.go
+++ b/weed/topology/topology_stats_test.go
@@ -111,3 +111,25 @@ func TestCollectionVolumeStatsWithEcVolumes(t *testing.T) {
t.Errorf("ec stats query should not create a phantom collection")
}
}
+
+func TestCollectionVolumeStatsClampsDeletedOverTotals(t *testing.T) {
+ topo := NewTopology("weedfs", sequence.NewMemorySequencer(), 32*1024, 5, false)
+
+ rack := topo.GetOrCreateDataCenter("dc1").GetOrCreateRack("rack1")
+ dn := rack.GetOrCreateDataNode("127.0.0.1", 34534, 0, "127.0.0.1", "", map[string]uint32{"": 25})
+
+ // Volume 1 reports more deleted than it holds, on both counters.
+ topo.SyncDataNodeRegistration([]*master_pb.VolumeInformationMessage{
+ {Id: 1, Size: 1000, FileCount: 10, DeletedByteCount: 4000, DeleteCount: 40, Version: uint32(needle.GetCurrentVersion())},
+ {Id: 2, Size: 3000, FileCount: 30, DeletedByteCount: 1000, DeleteCount: 10, Version: uint32(needle.GetCurrentVersion())},
+ }, dn)
+
+ // VolumeLocationList.Stats only counts nodes connected for over a minute
+ dn.LastSeen = time.Now().Unix() - 61
+
+ stats := topo.CollectionVolumeStats("")
+ // Volume 1 contributes nothing rather than wrapping.
+ assert(t, "used size", int(stats.UsedSize), 2000)
+ assert(t, "logical used size", int(stats.LogicalUsedSize), 2000)
+ assert(t, "file count", int(stats.FileCount), 20)
+}
diff --git a/weed/topology/volume_location_list.go b/weed/topology/volume_location_list.go
index be859d3d54..7fdbfd5564 100644
--- a/weed/topology/volume_location_list.go
+++ b/weed/topology/volume_location_list.go
@@ -96,13 +96,21 @@ func (dnll *VolumeLocationList) Refresh(freshThreshHold int64) {
}
}
-// Stats returns logic size and count
+// Stats returns logic size and count. Both subtractions are clamped: the
+// deleted counters are maintained apart from the totals they come off and can
+// transiently exceed them.
func (dnll *VolumeLocationList) Stats(vid needle.VolumeId, freshThreshHold int64) (size uint64, fileCount int) {
for _, dnl := range dnll.list {
if dnl.LastSeen < freshThreshHold {
vinfo, err := dnl.GetVolumesById(vid)
if err == nil {
- return (vinfo.Size - vinfo.DeletedByteCount), vinfo.FileCount - vinfo.DeleteCount
+ if vinfo.Size > vinfo.DeletedByteCount {
+ size = vinfo.Size - vinfo.DeletedByteCount
+ }
+ if vinfo.FileCount > vinfo.DeleteCount {
+ fileCount = vinfo.FileCount - vinfo.DeleteCount
+ }
+ return size, fileCount
}
}
}