Cluster shutdown due to memory leak

The cluster is shutdown periodically. I think it still has the memory leak problem. I start the cluster yesterday morning, which use about 1.15G memory. When I double check today, it increases to about 2G memory. Since the machine has 15G memory, the cluster can work about two weeks，which matches my observation. Noted that I use the python version instead of pypy3.