We just finished rolling out Tailscale to our whole support team of about 50 people. The goal was to securely access our internal helpdesk and CRM tools without using a clunky VPN.
Some things went really smoothly—installing it was easy for everyone. But we hit a few unexpected snags. A handful of users couldn't connect to certain resources, and we had some confusion around access controls. For those who've managed larger rollouts, what are the common pitfalls with ACLs or subnet routers? I'm also curious what best practices are for onboarding a group this size.
Congrats on getting 50 folks onboarded. That's a solid win right there.
The ACL confusion is super common in bigger groups. People often forget that the default "auto approve" rules can be too permissive. Did you start with a deny-by-default policy for your helpdesk subnet? That's what saved us headaches later.
For onboarding, I made a 5-minute Loom video showing how to check their connection status and access one key tool. Sent it with the install link. Cut our support tickets in half. Could something like that work for your CRM access issues?
That's a solid rollout size, congrats! The ACL confusion and those random users who can't connect is exactly where we stumbled too. I'd bet the access issues are a mix of tags and subnet router configs.
For your 50-person group, I found pre-creating the ACLs in a test environment with a couple of dummy users was a lifesaver. We used Terraform to manage the Tailscale ACLs (they have a provider!) so we could version control and review changes before they hit production. Something like this for a basic deny rule:
```hcl
resource "tailscale_acl" "policy" {
policy = jsonencode({
acls = [
{
action = "accept",
src = ["group:support"],
dst = ["helpdesk-subnet:*"],
}
]
})
}
```
Also, double-check that your subnet routers are advertising the correct routes. A few of our "can't connect" cases were just devices stuck on an older DNS server after switching from the old VPN.
Infrastructure as code is the only way