I am currently trying to manage my project budget on Google Cloud Platform and I’m confused about how billing works for Compute Engine. If I shut down my VM instance through the console so it is no longer running, will Google still charge me for the resources? I want to know if "stopping" an instance is enough to halt all costs, or if there are hidden fees for persistent disks and reserved IP addresses that I need to be aware of before leaving my lab environment idle for the weekend.
3 answers
When you stop a VM in GCP, you stop paying for the actual CPU and RAM (the compute power). However, you are definitely still charged for several other components that remain reserved for your account. The most significant cost is for the Persistent Disk (PD); since your data is still stored on that disk, you pay for its capacity as long as it exists. Additionally, if you have a Static External IP address assigned to the VM, GCP may charge a higher "unallocated" rate when the VM is stopped to discourage IP hoarding. To avoid all charges, you would need to delete the instance and its associated disks entirely.
That is a very common point of confusion for those new to Cloud Technology! Are you currently using a standard Persistent Disk or a more expensive SSD, and is your IP address ephemeral or static? These factors will determine exactly how much you see on your billing dashboard at the end of the month even if the machine is powered down.
You aren't charged for the instance itself, but you are charged for storage. Just remember that "Stopped" is different from "Terminated" in some other clouds; in GCP, stopping keeps your disk data safe but billable.
I agree with Lisa. It’s a vital distinction. As Kimberly mentioned, the "hidden" disk costs are what usually surprise people. It’s always best to check the GCP Pricing Calculator to see the specific cost per GB for your region.
Mark, I am using a 100GB Balanced Persistent Disk and I believe the IP is currently static. If I change the IP to ephemeral while the instance is stopped, will that help reduce the idle costs? I am trying to keep the VM configuration intact for a demo next month, but I don't want to spend my entire budget on a machine that isn't actually processing any data during the wait.