Arista EOS

Infrastructure Username and password (eAPI or SSH)

Check, upgrade and roll back Arista EOS switches over eAPI or SSH, directly from a Windmill flow.

Set up in 2 steps

1

Pick the transport

eAPI needs management api http-commands, then no shutdown, in EOS configuration mode. Where only SSH is enabled, set Transport to ssh on the resource: the actions and their results are the same. Use a user with privilege 15.

2

Add a resource

In Resources, add an arista_eos resource with the username and password. Set the switch's management address for a resource per switch, or leave it empty to share one resource between every switch with that login, and pass host to each action.

Scripts

Building blocks, if none of the projects above fit. Fork one, run it, or call it as a step inside your own flow.

Submit a script →

Resource type

The shape of the credential every script on this page expects.

FieldTypeRequiredWhat it is
usernamestringYesEOS user with privilege 15 (config sessions, boot system and reload need it).
passwordstringYesPassword of that EOS user.
transportstringNoHow the actions reach the switch: eapi (default) or ssh, for a switch where eAPI is not enabled. Same username and password, same actions and results.
portnumberNoPort of eAPI, or of SSH with transport ssh. Leave empty for the default: 443 for eAPI (80 for an http:// host), 22 for SSH.
verify_tlsbooleanNoVerify the switch's TLS certificate (transport eapi). Untick for the self-signed certificate EOS ships with.
ca_certificatestringNoWith transport eapi: the PEM certificate of the CA that signed the switch's eAPI certificate (-----BEGIN CERTIFICATE----- ...), for switches with a certificate from a private CA. It replaces the system's trusted CAs. Leave empty to use those.
allow_httpbooleanNoAllow an http:// host (eAPI's management api http-commands / protocol http). The password is then sent in cleartext: keep it off unless the switch is on an isolated management network.
ssh_host_keystringNoWith transport ssh: the switch's host key fingerprint as ssh-keygen -lf prints it (SHA256:..., the ED25519 key when the switch has one), checked on every connection. Leave empty to accept any host key.
hoststringNoSwitch management hostname or IP, e.g. leaf1.example.net or 10.0.0.11; with transport eapi a full URL (https://10.0.0.11:8443) also works. Leave empty for a resource shared by several switches: each action then takes host as an input, which also overrides this one.

Questions

The switch's certificate isn't trusted

If it comes from your own CA, paste that CA's PEM certificate in the resource's ca_certificate field. For the self-signed certificate EOS ships with, untick Verify TLS instead: the scripts then skip certificate verification for that switch.

How do I run an upgrade?

In one flow: Device pre-check, Device activate, Device reload, Get Reload Delay, Device post-check, then Device commit, or Device rollback when the post-check fails. Each script is one attempt: give Get Reload Delay a retry on DeviceUnreachable and a sleep of result.remaining_seconds, and Device post-check a retry on NotStable. Device pre-check returns its findings in problems without failing, so branch on it. MLAG is optional: without it, Get Reload Delay returns 0. Reload refuses unsaved config unless allow_unsaved is set.

Can I keep the upgrade logic in my own flow?

Yes: Get Facts, Get Flash, Get Boot Config, Copy Image, Set Boot Image, Reload and Delete File are single steps with no state between runs. The Device actions bundle one opinionated sequence (checks, snapshot comparison, reload once per run) on top of the same commands.

Can a switch event trigger a flow?

Not from this integration: schedule a flow that runs Get MLAG or Get Interfaces instead.