In the following, I divide the problem finding and problem solving into several stages.
stage 1. hanging on when accessing nfs
nfs can be exported in the nfs server and can be mounted in the nfs client.[https://wiki.gentoo.org/wiki/Nfs-utils]
But it will hang on when connectting the client through
ssh or accessing the nfs.
I still can excute command through the following method
1 ssh heq "ls /"
stage 2. exporting and mounting nfs through normal network instead of ib network
If nfs is exported in the server and mounted in the client through normal network instead of infiniband network, ssh connection is successful and nfs is available.
So, there is something wrong about infiniband network.
ssh through ib network
next, it was found that
ping works well through ib network. but ssh always fails through ib network.
Then the following command is used to find the root of the problem
1 2 3 4 ping -M do -s 100 ibio01 ping -M do -s 200 ibio01 ping -M do -s 300 ibio01 ping -M do -s 400 ibio01
it means the value of
mtu of infiniband switch is too small.
ssh connection stop at "debug1: SSH2_MSG_KEXINIT sent"
解决方案： ib 网卡的 mtu 值需要一致(destination 的不能太大)