diff --git a/lib/cmdlib.py b/lib/cmdlib.py
index e491163ac3182fed7cc4cdfc42d2a39df4c34ad1..87f145234d3119474878450846af40ccc162bc26 100644
--- a/lib/cmdlib.py
+++ b/lib/cmdlib.py
@@ -1608,6 +1608,7 @@ def _WaitForSync(lu, instance, oneshot=False, unlock=False):
     lu.cfg.SetDiskID(dev, node)
 
   retries = 0
+  degr_retries = 10 # in seconds, as we sleep 1 second each time
   while True:
     max_time = 0
     done = True
@@ -1640,6 +1641,16 @@ def _WaitForSync(lu, instance, oneshot=False, unlock=False):
           rem_time = "no time estimate"
         lu.proc.LogInfo("- device %s: %5.2f%% done, %s" %
                         (instance.disks[i].iv_name, perc_done, rem_time))
+
+    # if we're done but degraded, let's do a few small retries, to
+    # make sure we see a stable and not transient situation; therefore
+    # we force restart of the loop
+    if (done or oneshot) and cumul_degraded and degr_retries > 0:
+      logging.info("Degraded disks found, %d retries left", degr_retries)
+      degr_retries -= 1
+      time.sleep(1)
+      continue
+
     if done or oneshot:
       break