Waiting for servers after reboot

Archives Ansible Project
1

Hi ansible users,

I have a playbook to reboot all my servers serially, using the “serial: 1” option in the playbook. I have the tasks:

  • command: /sbin/reboot

  • local_action: wait_for host={{inventory_hostname}} port=22 state=stopped

  • local_action: wait_for host={{inventory_hostname}} port=22

However, some of my managed servers are not directly reachable from my laptop, so I usually connect to them via a bastion host. This means that the port 22 test will fail for them.

I thought about using the ping module instead, which should just do an SSH ping to a host. However, the ping module times out after a while when the underlying ssh connection times out, so I can’t make it wait long enough for the server to finish rebooting.

Does anyone have any ideas or suggestions on how to work around this?

Anand

2

“However, some of my managed servers are not directly reachable from my laptop, so I usually connect to them via a bastion host. This means that the port 22 test will fail for them.”

You could consider running Ansible from the bastion host, versus using an SSH jumphost which I infer you are using above.

The ping module doesn’t ICMP ping, BTW, it’s a basic Ansible connectivity test.

3

You could consider running Ansible from the bastion host, versus using an SSH jumphost which I infer you are using above.

Yes, by bastion, I meant an SSH jump host.

The ping module doesn’t ICMP ping, BTW, it’s a basic Ansible connectivity test.

Yes, I realise that the ansible ping module is not ICMP, but an SSH connection. The SSH jump host doesn’t have ansible installed, and I don’t have root access on it. I can of course install ansible in a virtualenv, but I was hoping for some cleverer way of detecting host down/up from within a playbook run from my laptop. I suppose my use-case is an outlier, so I’ll try to see what I can do on the jump host.

4

Yeah, I’m open to the idea of a ping module action_plugin that takes a ‘mode=icmp’ (default ssh).

You could also just do a “shell: ping” (making sure to ping just once) and register the return code.

5

i.e. local_action: ping …

6

I am running into a similar situation and have been thinking about using the shell: ping approach too. But what I came up in the following didn’t seem to do the trick. Would appreciate a hint:

7

Quick correction and follow-up:

I should have been more careful when cut-n-paste.

The task include file should be