# Control GCP costs with 22 workflows for FinOps > FinOps workflows to discover, approve, and automate GCP cost savings recommendations. By Turbot Team Published: 2024-07-24 Cloud infrastructure managers are always looking for ways to save costs. It's good to detect underused or unused resources, and even better to automatically streamline or remove them. Enter the [GCP Thrifty mod](https://github.com/turbot/flowpipe-mod-gcp-thrifty) for Flowpipe, a new tool that enables you automate cost optimization. Let's see how it works. ## Detect and release unused external IPs Here's a Steampipe query that finds unattached external IPs. ```sql > select name, location from gcp_compute_address +-----------+-------------+ | name | location | +-----------+-------------+ | ip-addr-2 | us-central1 | | ip-addr-1 | us-central1 | +-----------+-------------+ ``` We're paying for these, and if we don't need them we should release them. To do that, we'll run the `detect_and_correct_compute_addresses_if_unattached` pipeline in the GCP Thrifty mod. Per the [README](https://github.com/turbot/flowpipe-mod-gcp-thrifty?tab=readme-ov-file#gcp-thrifty-mod-for-flowpipe), the setup entails: - Cloning the mod - Visiting its directory and installing dependencies (`flowpipe mod install`) - Starting Steampipe as a service (`steampipe service start`) - Using `credential_import` to give Flowpipe access to the credentials in Steampipe's `gcp.spc` - Optionally, configuring variables that govern the mod's behavior. Let's run the pipeline. ``` $ flowpipe pipeline run detect_and_correct_compute_addresses_if_unattached --arg='approvers=[""]' ``` With no approvers, the pipeline runs without interruptions and reports that both addresses are unattached. ``` Detected unattached Compute address ip-addr-1 [us-central1/encoded-blend-333001]. Detected unattached Compute address ip-addr-2 [us-central1/encoded-blend-333001]. ``` Now let's take action on them. ## Wizard mode for interactive decisions In wizard mode, Flowpipe prompts you to decide whether to release each address. It's the default, so we can run the pipeline with no arguments. ``` flowpipe pipeline run detect_and_correct_compute_addresses_if_unattached ``` Now, Flowpipe pauses on each detected address, so you can decide whether to release it. We'll choose `Skip` and keep this one for now. ``` ┃ Detected unattached Compute address ip1 [us-central1/encoded-blend-333001]. ┃ > Skip ┃ Delete Compute Address ``` Note that this interaction can instead happen in Slack, MSTeams, or email by naming — in `approvers` — one or more notifiers configured with alternate [integrations](https://flowpipe.io/docs/reference/config-files/integration). Flowpipe proceeds to the next detected address. This time we'll choose `Delete Compute Address`; Flowpipe takes that action and reports the outcome. ``` Deleted Compute address ip-addr-2 [us-central1/encoded-blend-333001]. ``` Nice! ## Delete old compute snapshots automatically You won't want to automatically release IP addresses without checking each one, but you might want to automatically delete compute snapshots older than a threshold number of days. There's another pipeline for that, `detect_and_correct_compute_snapshots_exceeding_max_age`. To use that pipeline automatically, you'll activate a [query trigger](https://flowpipe.io/docs/flowpipe-hcl/trigger/query) to query for old snapshots. And you'll set some configuration variables to control how it works. We'll put them into a file, `continuous-compliance.fpvars`. ```hcl approvers = [] compute_snapshots_exceeding_max_age_trigger_enabled = true compute_snapshots_exceeding_max_age_trigger_schedule = "weekly" compute_snapshots_exceeding_max_age_default_action = "delete_snapshot" compute_snapshots_exceeding_max_age_days = 90 ``` Then we'll run Flowpipe in server mode and point it at that package of variables. ``` flowpipe server --var-file=continuous-compliance.fpvars ``` With `approvers = []` the pipeline handles each detected old snapshot without requiring approval. The default action is normally `notify` but here we override to `delete_snapshot`. Now, snapshots older than 90 days will be automatically deleted on a weekly basis. ## Configuration-driven workflow A detect-and-correct mod like this one doesn't require you to write any SQL or HCL. It's a tool that you configure and then use to perform a range of tasks you'd like to automate, with or without approval. Give GCP Thrifty a try, and please [let us know](https://turbot.com/community/join) how it goes. ## See it in action