test: remove flaky test from volumewatcher (#8189)

The volumewatcher restores itself on notification, but detecting this is racy
because it may reap any claim (or find there are no claims to reap) and
shutdown before we can test whether it's running. This appears to have become
flaky with a new version of golang. The other cases in this test case
sufficiently exercise the start/stop behavior of the volumewatcher, so remove
the flaky section.

nomad/volumewatcher/volumes_watcher_test.go

@@ -223,30 +223,6 @@ func TestVolumeWatch_StartStop(t *testing.T) {
 	require.Eventually(func() bool {
 		return !watcher.watchers[vol.ID+vol.Namespace].isRunning()
 	}, time.Second*5, 10*time.Millisecond)
-
-	// the watcher will have incremented the index so we need to make sure
-	// our inserts will trigger new events
-	index, _ = srv.State().LatestIndex()
-
-	// create a new claim
-	alloc3 := mock.Alloc()
-	alloc3.ClientStatus = structs.AllocClientStatusRunning
-	index++
-	err = srv.State().UpsertAllocs(index, []*structs.Allocation{alloc3})
-	require.NoError(err)
-	claim3 := &structs.CSIVolumeClaim{
-		AllocationID: alloc3.ID,
-		NodeID:       node.ID,
-		Mode:         structs.CSIVolumeClaimRelease,
-	}
-	index++
-	err = srv.State().CSIVolumeClaim(index, vol.Namespace, vol.ID, claim3)
-	require.NoError(err)
-
-	// a stopped watcher should restore itself on notification
-	require.Eventually(func() bool {
-		return watcher.watchers[vol.ID+vol.Namespace].isRunning()
-	}, time.Second*5, 10*time.Millisecond)
 }
 
 // TestVolumeWatch_RegisterDeregister tests the start and stop of