Run One-Click Remediation from a Recommendation

A cost recommendation is only a saving once someone acts on it. ZopNight can act for you on a defined set of findings through a Remediate button on the recommendation. This guide covers the prerequisites, the three kinds of recommendation you will meet, and how to run and check the first one.

Before the button works

The cloud account needs read + write access. Accounts connected read-only get discovery and recommendations, but every remediation, manual start or stop, and bulk action is blocked server-side before any cloud call. Change the permission level in the Edit Cloud Account drawer, and grant the matching cloud permissions on the provider side. An upgrade takes effect without a restart.

Your role decides whether you wait for an admin. Remediation runs one-click for admins. When a non-admin starts a destructive remediation, it pauses at an approval step until an admin approves it; see approving a destructive remediation.

Check the account’s permission report. View Permissions on the cloud account card shows each permission as Granted, Denied or Unknown, and ZopNight records per resource type whether the credential can discover, recommend or schedule it. A credential that can list a type but not change it will fail the action, so resolve denials first.

Three kinds of recommendation

Open any recommendation’s drawer and you will see one of these:

1. Auto. A Remediate button runs the fix end to end. Examples of the actions behind it: start or stop, pause a service, stop compute, resize (with a stop where the provider needs one), set a lifecycle policy, delete a resource, or patch a live Kubernetes manifest.
2. Guided. The same button, plus a review modal where you type to confirm before anything runs. Single-replica Kubernetes deployments are an example: scaling one replica to two is guided-only, because ZopNight will not change a workload the owner chose to run as a singleton without that confirmation.
3. Advisory. No button. The drawer shows a numbered “How to Fix” playbook to follow in your own console or pipeline, then you mark the recommendation Applied.

Schedule recommendations (compute that runs around the clock but is used in working hours) are a fourth path: their Remediate button opens schedule creation with the resource prefilled, and saving attaches it.

Run the first one

1. Pick a low-risk recommendation: an idle non-shared resource with a clear owner.
2. Open it and check the evidence block, which shows the monitoring values that triggered the rule.
3. Choose View Blast Radius to see connected resources and the risk score before you commit; see checking blast radius.
4. Choose Remediate. The wizard runs precondition, then approval if required, then the action, then validate.
5. Read the result. A failure is labelled user action (permission or quota, with a fix hint), transient (throttling or provider 5xx, retryable) or system (unsupported API).

Verify it outside ZopNight too

For the first few runs, confirm in the provider’s own tooling that the change landed, for example:

aws ec2 describe-instances --instance-ids i-0abc123def4567890 \
  --query "Reservations[].Instances[].[State.Name,InstanceType]" --output text

for a stop or resize on EC2. Over the following days, the recommendation’s history drawer records the status change and ZopNight re-checks the resource against the target it asked for, so a recommendation marked applied that the estate contradicts shows up as such.

What remediation does not touch

The pause template explicitly excludes customer databases (RDS, Aurora, Cloud SQL, ElastiCache, Azure SQL, PostgreSQL and MySQL types). Delete is in the action set, so treat a delete recommendation as permanent and back up first. Get notified of approvals, completions and failures by subscribing a channel to remediation events. Background: auto-remediation.