This message was deleted.
# general
a
This message was deleted.
r
I just retried the same SSH forward test from the machine that hosts rancher and it works equally well. That rules out firewall issues etc.
according to tcpdump on both the node and the host that rancher runs on, nothing is sent towards the node from rancher when it shows the "Failed to set up SSH tunneling for host" error
r
I think Rancher does some things via triggering from an encrypted, preexisting channel that it sets up with its agent on the cluster. I think general network rules are client clusters can get to Rancher but Rancher is not guaranteed to get to client clusters. So it probably does trigger that way.
(primarily saying for which direction to try wireshark)
r
there is no traffic either direction
🤷 1
r
No clue then, that's the only guess I had.
r
I'm capturing on "any" using these filter expressions:
Copy code
rancher side: '( tcp port 22 and host not <my workstation IP> ) or host <node IP>'
node side: '( tcp port 22 and host not <my workstation IP> ) or host <rancher IP>'
so basically i should get any SSH traffic going anywhere from/to rancher or the node except my workstation, as well as any traffic between rancher and the node
r
Only other guess I have is if the rancher & node are VMs on the same hypervisor, the traffic might not get to where libpcap is watching.
r
different ESXes
r
You can miss the occasional packet from paging when trying to capture on one of the machines involved in traffic, but all of them consistently seems unlikely. I'm assuming you get enough other traffic to know that libpcap is working and it's not something getting confused with network virtualization.
r
if i establish an SSH connection between the two hosts myself, I see that in both wiresharks
what manages rancher's encrypted tunnel on the node side? is that the kube-proxy container I see on the node? How do I tell if the tunnel is set up and working?
alright, made a post there, and I'm calling it a day