Optimization: switching off non-production OCI out of hours with Resource Scheduler

#The saving

Running weekdays 07:00–19:00 instead of 24×7 is 60 of 168 hours: roughly 60% less compute cost for those resources. Storage keeps billing.

#1. Tags first

Use tag defaults per compartment so new resources get the right environment tag.

#2. Resource Scheduler

Resource Scheduler creates schedules that run an action (start or stop) on selected resources at a recurrence (cron-style or iCal). Resources can be selected statically or dynamically by tags/compartment, so new dev resources are included automatically. Check the list of supported resource types (compute instances, instance pools, Autonomous Database and others are supported; the list grows).

Typical pair of schedules:

IAM: Resource Scheduler needs policies allowing it to manage the selected resources. Example shape (adjust to the documented resource principal syntax):

Allow any-user to manage instance-family in compartment dev where all {request.principal.type = 'resourceschedule'}
Allow any-user to manage autonomous-database-family in compartment dev where all {request.principal.type = 'resourceschedule'}

#3. Resources it doesn't cover: Functions

For anything not supported, use a scheduled Function (via Resource Scheduler invoking a function, or another scheduler) that reads tags and calls the API:

import oci, datetime

def stop_dev_instances(signer, compartment_id):
    compute = oci.core.ComputeClient(config={}, signer=signer)
    today = datetime.date.today().isoformat()
    for inst in oci.pagination.list_call_get_all_results(compute.list_instances, compartment_id,
                                                         lifecycle_state="RUNNING").data:
        ops = (inst.defined_tags or {}).get("Operations", {})
        if ops.get("Environment") not in ("dev", "test"):
            continue
        if ops.get("ScheduleOverrideUntil", "") >= today:
            continue
        compute.instance_action(inst.id, "SOFTSTOP")

Use the function's resource principal (oci.auth.signers.get_resource_principals_signer()) and a dynamic group with least-privilege policies.

#4. Per-resource notes

#5. Overrides and notifications

• Override tag with a maximum length (e.g. 14 days), checked by the function/automation.
• A message to the team channel 30 minutes before stop (Notifications topic → chat).
• Avoid stopping during running CI/CD deployments.

#6. Side effects

• Scheduled jobs inside dev VMs and databases: move them into working hours.
• Ephemeral public IPs change after stop/start. Use reserved IPs or DNS.
• Shared dev services used by other teams may need a longer window.

#7. Report savings

Monthly Cost Analysis for dev/test compartments, before and after, shared with the teams.

Tip

Pilot with one compartment for two weeks. Someone will always find a server that "must" stay on; the override tag handles it without weakening the schedule for everyone else.

#Checklist

• Environment and schedule tags with defaults
• Resource Scheduler stop/start schedules with dynamic selection
• IAM policies for the scheduler and functions
• Functions for unsupported resources
• Override tag and pre-stop notification
• Side effects checked; savings reported