I once had a problem with an instance running on Openstack, namely that it could not ssh to other nodes, but other nodes could ssh to the said node. It turns out that this is because the MTU is different at the instance or at the server host level, please check the MTU again on each interface related to openstack's neutron external.
ssh -vvv -p $MUDDYPORT $MUDDYIP
- hangs at
expecting SSH2_MSG_KEX_ECDH_REPLY
ping node2 -c 8000
- If result RTO, this is cause MTU not set properly on Instance or Sever Nodes.
- Please adjust `8000` to our enviroment MTU.
- In my case i use 8950 MTU for Instance & 9000 for Server Host level.
$ sudo ifconfig enp71s0 mtu 9000
$ ssh -p $MUDDYPORT $MUDDYIP
...
thomasn@muddy:~$
## Verify
$ ping -M do $MUDDYIP -s 8000
PING 81.155.215.123 (81.155.215.123) 8000(8028) bytes of data.
1208 bytes from 81.155.215.123: icmp_seq=1 ttl=50 time=46.1 ms
1208 bytes from 81.155.215.123: icmp_seq=2 ttl=50 time=46.5 ms
$ ping -M do $MUDDYIP -s 8000
...
From 10.10.10.31: icmp_seq=1 Frag needed and DF set (mtu = 1500)
From 10.10.10.31: icmp_seq=1 Frag needed and DF set (mtu = 1500)^C
--- 81.155.215.123 ping statistics ---
0 packets transmitted, 0 received, +92331 errors
- Discussion
- Fix: Adjusting MTU