Dist.init_process_group works but TCPStore failed

Hi,
I am trying init dist and get stuck.
I have 2 nodes: master and slave, both pytorch 1.3.1 installed by anaconda
It works on both when:

    dist.init_process_group(
            backend ="NCCL",
            world_size = 2,
            rank = 0,# 0 for master and 1 for slave
            init_method="tcp://192.168.1.102:23458"#master addr and port
            )

It is hung up on both when:

store = dist.TCPStore("192.168.1.102", 23458, 2, 0)

Could somebody help?
Thanks in advance and Happy New Year.

This is how TCPStore is initialized when you call init_process_group. You can print out the args on both process to check which one went wrong.

For start_daemon, did you pass in 1 for rank0 and 0 for rank1?