Hacker News new | ask | show | jobs
by dvno42 1087 days ago
Hey this is cool! Thanks for sharing your hard work.

I have been living this for the past few years building an automation product[0] and services company to lower the barrier of entry and have tested many of these methodologies. We’ve also written many different runbooks/playbooks for complicated workflows. I’d like to share a couple experiences/opinions:

Netconf and vendor apis are lovely when available and working well. Many devices don’t support this and falling back to SSH (sometimes even telnet) is a must for automation. Imo, you could add value to your book by touching on Ktbyer's Netmiko/Paramiko[1] as well as their nuances (timeouts, dealing with interactive prompts, etc).

AAA is a big component of automation too. Having something in place to handle authn/authz (radius/tacacs) enables consistency for access across vendors. This also enables least privileged accounts and rotation/limited lifetime of creds when used with something like Hashicorp Vault[2]. I think you briefly mentioned secrets management though Ansible vault.

Another technology that may be worth mentioning is Textfsm[3] in conjunction with Netmiko. When we automate workflows for clients, there’s often times where the data we need to parse isn’t easily parsable. Using and expanding on textfsm makes this doable.

Lastly, much automation may only be one firmware change away from breaking. Even with the big vendors, bugs are common that are (ime) low priority to the OEM. Keep this in mind when writing runbooks/playbooks, try to rely on features and output that are unlikely to change across versions.

[0]https://realmhelm.com [1]https://github.com/ktbyers/netmiko [2]https://github.com/hashicorp/vault [3]https://github.com/google/textfsm
1 comments
Cyph0n 1087 days ago
+1 to textfsm: it is an extremely powerful approach to reliably parse CLI-based outputs. I used to do some IOS-XR device automation when I worked at Cisco - mainly for integration testing - and I (and other teams) used it heavily.

This ties in to your point about how you often need to fallback to SSH or Telnet. For example, a lot of platform-specific data isn’t exposed through standard interfaces, but almost everything is available through a CLI. There are also times when you have no choice but to use the CLI - for example, when re-imaging or reloading a device.

link