Lightweight server monitoring with historical data, docker stats, and alerts.
fix(agent): don't fail degraded RAID1 arrays that still have redundancy (#2760)
yi111ed025bc438e5
diff --git a/agent/mdraid_linux.go b/agent/mdraid_linux.goindex b56ec6221..4f2d71054 100644--- a/agent/mdraid_linux.go+++ b/agent/mdraid_linux.go@@ -164,9 +164,10 @@ func mdraidSmartStatus(health mdraidHealth) string {return "FAILED"}if health.degraded > 0 {- if isSparseSlotDegraded(health) {+ if isSparseSlotDegraded(health) || isRedundantMirrorDegraded(health) {// A sysfs snapshot cannot distinguish reserved slots from a removed- // member on sparse arrays, so report the ambiguity as a warning.+ // member on sparse arrays, and a mirror with two or more active+ // members is still redundant, so report these as a warning.return "WARNING"}return "FAILED"@@ -221,6 +222,19 @@ func isSparseSlotDegraded(health mdraidHealth) bool {return sparseSlots > health.populatedDisks}+// isRedundantMirrorDegraded reports whether a degraded RAID1 array still has at+// least two active members, so every block still has a redundant copy. Synology+// DSM sizes its system arrays (md0/md1) to the bay count, leaving unpopulated+// bays as missing mirror slots; whether those were never used or a member was+// lost, the remaining mirrors keep the array redundant.+func isRedundantMirrorDegraded(health mdraidHealth) bool {+ if !strings.EqualFold(strings.TrimSpace(health.level), "raid1") || health.degraded >= health.raidDisks {+ return false+ }+ activeDisks := health.raidDisks - health.degraded+ return activeDisks >= 2+}+// isMdraidBlockName matches /dev/mdN-style block device names.func isMdraidBlockName(name string) bool {if !strings.HasPrefix(name, "md") {diff --git a/agent/mdraid_linux_test.go b/agent/mdraid_linux_test.goindex 93572be19..2af73ad64 100644--- a/agent/mdraid_linux_test.go+++ b/agent/mdraid_linux_test.go@@ -148,6 +148,26 @@ func TestMdraidSmartStatus(t *testing.T) {if got := mdraidSmartStatus(mdraidHealth{arrayState: "clean", degraded: 1, faultyDisks: 0, raidDisks: 4, populatedDisks: 3}); got != "FAILED" {t.Fatalf("mdraidSmartStatus(missing member) = %q, want FAILED", got)}+ // Synology-style: RAID1 system array sized to 8 bays with 5 drives installed.+ if got := mdraidSmartStatus(mdraidHealth{level: "raid1", arrayState: "clean", degraded: 3, faultyDisks: 0, raidDisks: 8, populatedDisks: 5}); got != "WARNING" {+ t.Fatalf("mdraidSmartStatus(synology raid1 sparse bays) = %q, want WARNING", got)+ }+ // A 3-way mirror missing one member still has two copies.+ if got := mdraidSmartStatus(mdraidHealth{level: "raid1", arrayState: "clean", degraded: 1, faultyDisks: 0, raidDisks: 3, populatedDisks: 2}); got != "WARNING" {+ t.Fatalf("mdraidSmartStatus(3-way mirror missing member) = %q, want WARNING", got)+ }+ // A 2-disk mirror missing one member has no redundancy left.+ if got := mdraidSmartStatus(mdraidHealth{level: "raid1", arrayState: "clean", degraded: 1, faultyDisks: 0, raidDisks: 2, populatedDisks: 1}); got != "FAILED" {+ t.Fatalf("mdraidSmartStatus(2-disk mirror missing member) = %q, want FAILED", got)+ }+ // A faulty member still fails a mirror even with other members active.+ if got := mdraidSmartStatus(mdraidHealth{level: "raid1", arrayState: "clean", degraded: 1, faultyDisks: 1, raidDisks: 3, populatedDisks: 3}); got != "FAILED" {+ t.Fatalf("mdraidSmartStatus(mirror with faulty member) = %q, want FAILED", got)+ }+ // Parity arrays missing a member still fail.+ if got := mdraidSmartStatus(mdraidHealth{level: "raid6", arrayState: "clean", degraded: 1, faultyDisks: 0, raidDisks: 5, populatedDisks: 4}); got != "FAILED" {+ t.Fatalf("mdraidSmartStatus(raid6 missing member) = %q, want FAILED", got)+ }// Degraded with no member-state info at all (e.g. sysfs read failed) must// still fail rather than being silently treated as a sparse QNAP array.if got := mdraidSmartStatus(mdraidHealth{arrayState: "clean", degraded: 1, faultyDisks: 0, raidDisks: 4, populatedDisks: 0}); got != "FAILED" {