Costs and savings

Railyard never resells compute: your servers are billed by your cloud provider at their price. The Costs page (sidebar → Operate → Costs) reads those bills and the resources in your accounts, and shows where money leaks.

Run a scan

Costs uses the cloud keys under Settings → Team → Cloud credentials, the same keys you use to create servers. Press Scan now (admins). The scan runs in the background with live progress: each account in turn, AWS region by region, and the resources found. You can leave the page while it runs.

Accounts that can be scanned: DigitalOcean, Hetzner, Vultr and AWS. Servers on other providers, and servers you connected over SSH, aren’t part of the scan.

What you get

• Billed amounts straight from each provider’s billing API: AWS Cost Explorer, DigitalOcean invoices and Vultr pending charges, including credits and refunds. Hetzner has no billing API, so its figures come from Hetzner’s live price list.
• Railyard servers with their provider, size, apps, and 14-day average and peak CPU and memory.
• Everything in your accounts at list price: servers, disks, IP addresses, snapshots, machine images, NAT gateways, load balancers and databases, including things Railyard didn’t create.
• Savings, grouped as Easy fixes, Medium, Bigger changes and Worth a look. Each one says why, what fixing it does to your app, and how.

Findings include:

• unattached disks, and disks on long-stopped servers;
• unused IP addresses;
• NAT gateways and load balancers with no traffic;
• old snapshots and unused machine images;
• idle servers, leftover servers, and servers with no apps or only stopped apps;
• databases with no connections;
• a smaller size that fits the load;
• previous-generation AWS instances;
• ARM (Graviton) servers for less;
• AWS disks and database storage that could move to gp3;
• folding one server into another;
• AWS Savings Plan recommendations;
• spending spikes;
• data transfer taking a large share of the bill.

Fix it in one click

Where the fix is safe to automate, the finding has a button: Delete disk, Release IP, Delete snapshot, Terminate, Convert to gp3, Stop server or Resize. Every button asks you to confirm and tells you the impact first, for example “Deletes the disk and its data — snapshot it first”.

A resize takes a server snapshot first. Bigger changes, like moving to ARM, come with steps instead of a button: add the new server, then use Settings → General → Move to another server on each app.

Snooze 30d or Ignore hides a finding. Ignored findings stay listed at the bottom, where you can restore them.

Budget and alerts

Set a Monthly budget on the page. Rules with costs ticked under Team → Notifications get a message at 50%, 80% and 100% of the month-end forecast against that budget, plus whenever spending spikes or new waste appears.

Tick Hard cap to have Railyard act at 100%, not just alert: it puts every non-production app to sleep and blocks creating new servers until you’re back under budget or raise it.

AWS permissions

For AWS, the page shows the exact IAM policy to attach. It has two parts:

• Read-only access to Cost Explorer, pricing, EC2, load balancers, RDS and CloudWatch.
• Only the actions the one-click fixes need, if you want the buttons: delete volume, release address, delete snapshot, modify volume, stop and terminate instances.

AWS charges a small fee per Cost Explorer request. Railyard makes a handful per refresh and refreshes billing at most every 12 hours.

Limits

• Savings figures use each provider’s list price for the region. Discounts you negotiated aren’t known.
• Costs per environment on the canvas are each service’s share of its server’s monthly price, worked out from memory. They’re an estimate, not a bill.