tunnel_check_sleep behavior is not straightforward

[sashasimkin]

Hi!

While using this module with SSM gateway on a slow connection I had intermittent "Connection refused" errors, while sometimes it did work, when the speed of internet worsened it completely blocked my terraform applys.

While reading the inputs to the module, I saw tunnel_check_sleep, that could help in my situation, however when I set it to 5, then 10, then 30 seconds nothing changed in the terraform logs, no longer delay and I was getting same "Connection refused" errors.
The weird thing here was that it was still taking 3 seconds to "read" the data.external.ssh_tunnel resource.

After that, I tried setting parent_wait_sleep to a bigger value (10), and errors went away, and I also saw that now it took 10 seconds to retrieve data.external.ssh_tunnel.

When I went to the source code I saw the logic of parent/child tunnel initiation, and specifically that this is how actual tunnel is established:

nohup timeout "$TUNNEL_TIMEOUT" "$TUNNEL_SHELL_CMD" "$TUNNEL_ABSPATH/tunnel.sh" "$p" <&- >&- 2>"$clog" &

which means that it's running in the background still, but also, terraform doesn't wait for childs to return control back to the other resources.

While writing this and looking at the code I just realized that the parent_wait_sleep should be > tunnel_check_sleep for the encoded logic to work.
From the documentation, however it reads that they're additive and not overlapping.

I'm not sure if this is something that you would consider as worth to fix, but I think it makes sense to document more the relation of these two variables.