Features / Jump host

Jump hosts, the way ssh ‑J does it.

The machine you work on is often not the one you can reach: a GPU node behind a cluster’s login node, a server behind a bastion. Moshpit goes through the first to get to the second.

$ ssh -J login.example.com gpu01.example.com

Connection          gpu01.example.com
Jump Host › Connect via   login.example.com

Why it’s there

No shell inside a shell

Without it you connect to the login node and type ssh again inside the terminal. Moshpit then knows nothing about the second host: no panes, no agent state, no pushes from it.

Each host keeps its own keys

The jump host signs in with the key or password saved on its own connection; the target uses its own. Nothing is copied onto the jump host.

Failures say which host

If the connection fails, the card on Home names the host that didn’t answer and the reason.

How to use it

  1. Save the jump host as an ordinary connection, and check that it connects on its own.
  2. Open the target’s connection form. Under Jump Host, tap Connect via and pick the jump host.
  3. If the jump host is only a way in, turn on Jump host only: it stays off Home and still shows up in Connect via.

Good to know

  • One hop per connection. A chain of two or more jump hosts is not supported.
  • Through a jump host the connection uses SSH. Mosh needs UDP straight to the host.

Jump hosts in the docs →

More features