have you checked your cluster logs? system logs? does it always happen? have you investigated a bit by yourself?