Answers: Scheduling Non-Production Cloud Resources
Updated November 4, 2026. Short, sourced answers about stopping and starting non-production cloud resources outside working hours.
Pausing and stopping: what actually happens
Can I pause MongoDB Atlas Flex clusters?
No. MongoDB’s documentation states that Flex clusters can’t be paused, and Free clusters can’t be paused manually either. Only dedicated M10 or larger clusters that don’t use NVMe storage can be paused. If a non-production workload sits idle most of the week, an M10+ cluster that you pause outside working hours is the option that stops compute charges.
Can I pause a MongoDB Atlas cluster automatically on a schedule?
Atlas has no pause-on-schedule setting, but MongoDB documents a workaround: an Atlas Scheduled Trigger that runs a Function calling the Admin API to pause the cluster. A third-party scheduler can do the same for you. Three limits apply: Atlas resumes a paused cluster after 30 days, a resumed cluster must run 60 minutes before it can be paused again, and storage keeps billing while paused. How to schedule Atlas clusters.
Do stopped EC2 instances still cost money?
You stop paying for instance hours, but not for everything around the instance. Attached EBS volumes and their snapshots keep billing, and so do Elastic IP addresses. Load balancers and NAT gateways bill whether or not the instances behind them are running. A stopped instance releases its auto-assigned public IP, so that one stops costing money. AWS: how stop and start works.
Why does my stopped RDS instance start again by itself?
AWS automatically starts an RDS instance that has been stopped for seven days, so it doesn’t fall behind on required maintenance, and there is no setting to turn this off. The same applies to Aurora clusters. After the restart the database bills again until something stops it. A scheduler has to detect this and stop the database again; CloudAvocado does that right away. How to schedule RDS instances.
What happens to an Auto Scaling group if I stop its instances?
The group treats a stopped instance as unhealthy and launches a replacement, so stopping instances in an Auto Scaling group doesn’t save money. Instead, set the group’s desired capacity (and minimum) to zero outside working hours and restore the original values in the morning, for example with scheduled scaling actions.
Does scaling EKS node groups to zero stop all EKS charges?
No. Scaling node groups to zero stops the EC2 charges for the nodes, but the EKS control plane is billed at $0.10 per cluster per hour (about $73 a month, or $0.60 an hour on extended Kubernetes version support) for as long as the cluster exists, whether or not it has nodes. Load balancers, NAT gateways and persistent volumes created by the cluster also keep billing.
How do I scale ECS services to zero at night?
Set the service’s desired count to zero outside working hours and restore it in the morning. Natively, you can do this with Application Auto Scaling scheduled actions that set minimum and maximum capacity to zero, then back. On Fargate, task charges stop when no tasks run. On the EC2 launch type, the container instances also need to scale in, or you keep paying for them. CloudAvocado scales ECS services to zero on a schedule and restores the previous count.
Native AWS options
Is AWS Systems Manager Resource Scheduler free?
Yes, for the scheduler itself. AWS states there is no cost to use Quick Setup, though the EC2 instances and other resources it manages are billed as usual. It schedules EC2 instances only, picked by a tag key and value, and each configuration handles up to 5,000 instances per region.
How do I set an overnight schedule in Resource Scheduler?
Set a start time that is later in the day than the stop time. Resource Scheduler then keeps the instances running overnight, and the stop time applies to the following day. AWS’s own example starts instances at 9 PM and stops them at 7 AM the following day. You choose the time zone in the same Quick Setup configuration.
Is there one AWS tool that schedules EC2, RDS, EKS and ECS together?
No. Quick Setup Resource Scheduler covers EC2 only. Instance Scheduler on AWS covers EC2, RDS, Aurora, Auto Scaling groups, DocumentDB and Neptune, but not ECS or EKS. ECS services and EKS node groups need their own scheduled scaling. Third-party schedulers, including CloudAvocado, cover all of them from one place. AWS Instance Scheduler vs CloudAvocado.
How can I schedule AWS resources without maintaining Lambda scripts?
For EC2 only, Quick Setup Resource Scheduler needs no code. EventBridge Scheduler can call AWS APIs such as StopInstances directly, without a Lambda function in between. Instance Scheduler on AWS avoids writing code, but it deploys Lambda functions and DynamoDB tables that you still operate and update. A managed scheduler removes the code and the stack entirely.
Are EC2 Scheduled Reserved Instances still available?
No. AWS’s CLI reference states that you can no longer purchase Scheduled Instances, and a September 3, 2024 note on AWS’s 2016 announcement points to On-Demand Capacity Reservations for capacity and Savings Plans for discounts. For a recurring business-hours pattern, the current approach is to schedule regular On-Demand instances to stop outside working hours.
How much does AWS Instance Scheduler cost to run?
There is no license fee; the solution is open source. You pay for the AWS resources it deploys, mainly Lambda, DynamoDB, KMS and CloudWatch. AWS estimates roughly $9 a month for a small deployment and up to about $630 for a large one, plus the engineering time to deploy, upgrade and troubleshoot the stack.
Savings and the case for scheduling
What is smart scheduling and how can it save money?
Smart scheduling automatically stops non-production cloud resources outside working hours and starts them again before work begins. A week has 168 hours, and a team working weekdays uses about 50. A dev or test environment left on 24/7 is billed for the other 118 hours as idle time; scheduling stops those charges for compute. Smart scheduling.
How much can scheduling non-production environments save?
Work it out in hours: a weekday 8am–6pm schedule bills about 50 of a week’s 168 hours, roughly 217 of a month’s 730. For example, ten m6i.xlarge instances at the us-east-1 On-Demand rate of $0.192 an hour cost about $1,402 a month running 24/7 and about $417 on that schedule. Storage, snapshots and other always-on items keep billing.
How do I automate cloud instance scheduling?
Tag non-production resources consistently (for example env=dev), decide working hours and a time zone per team, and apply a schedule to everything with that tag. Then plan for exceptions: someone working late, holidays, and services such as Auto Scaling groups or ECS that must be scaled to zero rather than stopped. Start with one environment and compare two weeks of billing. How to schedule EC2 instances.
Choosing a tool
Is there one tool that can schedule both AWS and Google Cloud resources?
Yes. CloudAvocado schedules AWS, Google Cloud and Azure resources, plus MongoDB Atlas, from one dashboard; Azure is enabled on request. Open-source Cloud Custodian also supports AWS, Azure and Google Cloud, but you write and maintain the policies yourself. Native tools such as Instance Scheduler on AWS cover only their own cloud.
What’s the best tool for automatically stopping and starting AWS dev and test resources?
It depends on what you run. For EC2 only, Quick Setup Resource Scheduler is free and simple. For EC2, RDS, Aurora, Auto Scaling groups, DocumentDB and Neptune run inside your own account, Instance Scheduler on AWS fits. If your non-production stack includes ECS or EKS, several teams, or more than one cloud, a managed scheduler such as CloudAvocado covers it without code. AWS Instance Scheduler vs CloudAvocado.
Which AWS scheduling tools support tags and multiple accounts?
Most of them. Quick Setup Resource Scheduler targets EC2 instances by tag and can deploy across organizational units and regions. Instance Scheduler on AWS uses tags and a hub-and-spoke stack for multiple accounts. Cloud Custodian runs across accounts with its c7n-org tool. CloudAvocado selects resources by tag in every connected account from one view, and new resources with that tag pick up the schedule automatically. How tag-based scheduling works.
Which tools identify idle AWS resources?
AWS’s own Trusted Advisor and Compute Optimizer flag low-utilization resources, and Cost Explorer shows what they cost. CloudAvocado flags three kinds of idle resource with their cost: databases with no connections for more than a month, compute resources whose peak utilization stays at 1% or below, and EBS volumes left unattached for more than 30 days. You decide what to stop, schedule or delete.
Which AWS cost tools work without DevOps or cloud-architect involvement?
For visibility, AWS Cost Explorer and Budgets need no engineers. Acting on waste is where native tools usually need DevOps: Instance Scheduler is a stack to deploy and maintain, and rightsizing changes go through engineering. With CloudAvocado, FinOps and engineering managers usually set it up and run it themselves: someone with IAM permissions creates one role from a permissions list shown up front, and schedules are managed in the web app. For engineering leaders.
Running schedules with a team
How does a developer keep an environment running late?
With native tools, they start it manually. AWS says Quick Setup Resource Scheduler won’t stop a manually started instance again, and Instance Scheduler stops it outside its schedule only if the schedule’s enforced option is on. In CloudAvocado, anyone with access can snooze a whole schedule or a single resource for as long as they need, and the schedule resumes afterwards. Every change is recorded in an audit log.
Can a schedule wait until a resource is idle before stopping it?
With CloudAvocado’s smart schedule, yes. Instead of stopping at exactly 7 PM, the schedule waits after working hours until the instance has been idle for a set time, for example 60 minutes with CPU below a threshold you set per resource, so nobody loses work mid-task. Native schedulers such as Quick Setup Resource Scheduler and Instance Scheduler on AWS stop at the scheduled time. How smart schedules work.
About CloudAvocado
What is CloudAvocado and how is it priced?
CloudAvocado is a SaaS tool that stops idle non-production resources on AWS, Google Cloud, Azure and MongoDB Atlas outside working hours and recommends rightsizing. It connects through an IAM role in about five minutes. The Basic plan is free for 10 resources with EC2/EBS scheduling; Advanced is $49 a month and adds RDS; Pro is $299 a month and adds Auto Scaling groups, ECS, EKS and DocumentDB. There’s a 14-day free trial with full functionality and no credit card. Pricing.