# Destroying an environment

When an application's environment deploys to a
[cloud account](/user-guide/settings/workspace-settings/cloud-accounts), its stack runs there: the
app, its API, its database and its address. **Destroy** removes that stack through the repository's
destroy workflow, keeping a snapshot of the database first. The environment stays attached, so you can
deploy it again later, or detach it once its stack is gone.

:::note
Destroying environments is part of Cloud accounts, which is in early access.
:::

<Availability rollout permission="Anyone who can open the application sees an environment's destroy. Destroy, Cancel destroy and Try destroy again: the Owner and Admin roles, signed in to Productmind themselves, whatever your plan and the rollout. Deploy again also needs Cloud accounts in your plan and turned on for your organization." />

## Key concepts

**Destroy.** Productmind stops the environment's deploys, then runs the destroy workflow of the
application's repository in the environment's cloud account. It takes a snapshot of the database
first, when the stack has one, then removes the app, the API, their data and the address. It takes
about half an hour.

**The production wait.** A **Production** environment's destroy waits 24 hours before it starts.
Productmind tells the workspace's Owners and Admins, any of them can cancel it until then, and the
environment keeps deploying meanwhile. Any other environment's destroy starts at once.

**What stays.** A destroy keeps:

- the database's snapshot, when the stack has a database;
- in a **Production** environment, the uploads bucket with its files;
- the deploy state and artifacts buckets and their key (the key about USD 1 a month);
- in an account Productmind manages, the account's network, about USD 38 a month plus its public
  address, billed with the storage of what stays until the account is deleted.

In your own AWS account, everything that stays is in that account. The repository keeps the
environment and its settings, and the branch keeps its protection. A domain you own keeps its DNS
zone and the DNS records you keep in Productmind, so its mail keeps working.

**Deploy again.** A destroyed environment stays attached, and nothing deploys to it. **Deploy again**
builds it anew from its branch's newest commit, in the same account and at the same address. Its
database starts empty: the snapshot of the destroyed one stays in the account.

Each environment's row shows its destroy:

| State | What it means |
| --- | --- |
| **Destroy scheduled** | A **Production** environment's destroy waits its 24 hours. The row says when it starts, and the environment keeps deploying until then |
| **Destroying** | Deploys are stopped and the destroy has started |
| **Restoring** | Productmind is putting the environment back in its deploys, after a cancel or **Deploy again** |
| **Destroy failed** | The destroy stopped. The row says why and what to do, and nothing deploys until you try again, cancel or detach it |
| **Destroyed** | The stack was removed. The row says when, and what stays |

## The destroy at a glance

<Screenshot id="applications-cloud-accounts-destroy-states" alt="The Cloud accounts section of an application's Source code tab: Development, Destroyed, its stack removed, merges to main not deploying until Deploy again, and what stays in the account, with a Deploy again button; and Production, Destroy scheduled, destroying on a date a day away unless cancelled and deploying until then, with a Cancel destroy button" />

**Destroy...** is in the environment's row menu (the **...**), above **Detach...**, while the environment
is **Active**. Once a destroy is asked for, the row shows its state, with **Cancel destroy**, **Try
destroy again** or **Deploy again** where they apply.

## Destroying an environment

### To destroy an environment

1. Open the application and choose **Source code**.
2. Under **Cloud accounts**, choose the **...** on the environment's row, then **Destroy...**.
3. Read what Productmind does and what stays in the account.
4. Enter the environment's key, for example **development**, exactly as shown.
5. If you have set up an authenticator app, enter a code from it, or one of your recovery codes.
6. Choose **Destroy**, or **Schedule destroy** for a **Production** environment.

<Screenshot id="applications-cloud-accounts-destroy" alt="The Destroy Development of SupportHub dialog: Productmind stops deploys to Development, then runs the repository's destroy workflow in AWS account 210987654321, taking a snapshot of the database first; Development stays attached; what stays in the account and its cost; then a field to type development to confirm, a field for a code from your authenticator app or a recovery code, and the Cancel and Destroy buttons" />

The row shows **Destroying**, then **Destroyed**. A **Production** environment's row shows **Destroy
scheduled** and the time its destroy starts, then **Destroying**.

### To destroy an environment, then detach it

When an environment's stack may still run in its account, **Detach** asks first what happens to it.

1. Open the application and choose **Source code**.
2. Under **Cloud accounts**, choose the **...** on the environment's row, then **Detach...**.
3. Choose **Destroy, then detach**, then **Continue**.
4. Confirm the destroy as in [To destroy an environment](#to-destroy-an-environment).

Once its stack is removed, Productmind detaches the environment. To keep the stack running instead,
choose **Detach and keep it running**: see
[Cloud accounts](/user-guide/settings/workspace-settings/cloud-accounts#to-stop-an-environment-deploying-to-an-account).

### To cancel a destroy

1. Open the application and choose **Source code**.
2. Under **Cloud accounts**, choose **Cancel destroy** on the environment's row.
3. Read what the cancel does, then choose **Cancel destroy**.

A scheduled destroy has removed nothing: the environment keeps deploying. A destroy under way stops
and the environment goes back to deploying, unless its destroy job has already started: a destroy that
has started can't be stopped. After a destroy that failed while changing the stack, Productmind also
deploys the environment's branch to put back what the destroy changed. If the database was removed,
the new one is empty: its data is in the snapshot.

### To try a failed destroy again

1. Open the application and choose **Source code**.
2. Under **Cloud accounts**, choose **Try destroy again** on the environment's row.
3. Enter the environment's key and, if you have an authenticator app, a code from it.
4. Choose **Destroy**.

The destroy runs at once: a **Production** environment doesn't wait another 24 hours. Productmind
reads the repository's destroy workflow again, so a changed workflow is the one that runs.

### To deploy a destroyed environment again

1. Open the application and choose **Source code**.
2. Under **Cloud accounts**, choose **Deploy again** on the environment's row.
3. Choose **Deploy again**.

The row shows **Restoring**, then **Active**, and its deploy appears under **Deploys**: see
[Deploys](/user-guide/settings/workspace-settings/deploys).

## How destroys behave

**A destroy asks you to confirm it.** You type the environment's key, which proves which environment
you mean. If you have set up an authenticator app, the destroy also takes a fresh code from it or a
recovery code, which a stolen session can't give. A recovery code is used up once it is accepted: a
destroy refused before your code is checked keeps it. Wrong codes count toward your account's limit,
the same one as signing in.

**What stops while a destroy is asked for.** The environment's stack settings, its branch and its
domain can't change while a destroy is scheduled, under way, restoring or failed. Once it is destroyed,
its stack settings can change again and apply at **Deploy again**; its branch and its domain change
once it is deployed again or detached. **Retry deploy** and **Deploy now** aren't offered from the
moment the destroy starts until the environment deploys again. To detach it while its destroy is
scheduled or under way, cancel the destroy first; once the row shows **Active**, detach it, or destroy it
again choosing **Destroy, then detach**. A destroy whose job has started can't be cancelled: detach the
environment once the destroy ends.

**Who is notified.** The person who asked and the workspace's Owners and Admins are told at each step:
**Destroy scheduled**, **Destroy started**, **Environment destroyed** (with what stays),
**Destroy failed** and **Destroy cancelled**; and when GitHub refuses a step Productmind takes for them,
**Deploy not started**, **Cancel didn't go through** or **Deploy again didn't start**, each saying what
to do. Each appears in the app and is emailed, or arrives as a
Slack direct message instead when you have connected your Slack account (see
[How notifications reach you](/user-guide/profile-and-notifications#how-notifications-reach-you)).

**The rows keep themselves current.** While an environment is **Destroying** or **Restoring**, its
application's **Cloud accounts** rows refresh every 15 seconds.

**Deploys show the destroy.** The destroy's run appears in the environment's
[deploys](/user-guide/settings/workspace-settings/deploys), as **Destroy by** the person who asked.

**Each destroy is recorded.** Each destroy and each try again is recorded in the audit log, with who
asked, as is each change of the destroy's state.

## Permissions

| To | You need |
| --- | --- |
| See an environment's destroy and its state | Anyone who can open the application |
| Destroy an environment, cancel its destroy and try a failed destroy again | The **Owner** or **Admin** role, signed in to Productmind yourself |
| Deploy a destroyed environment again | The **Owner** or **Admin** role, signed in to Productmind yourself, and Cloud accounts in your plan |

Neither your plan nor the Cloud accounts rollout ever holds back destroying, cancelling or trying again.
Only **Deploy again**, which builds an environment anew, needs Cloud accounts in your plan and turned on
for your organization.

## Limits

- A **Production** environment's destroy waits 24 hours before it starts.
- A destroy that has started can't be stopped.
- Productmind starts a destroy at most three times in all: when newer runs replace it before it
  starts, or its request runs out first, it starts it again. After the third time it fails, and you can
  try it again.

## Troubleshooting

### "That code is not right."

The code was wrong, or a code from your authenticator app was already used. Enter a new code, or one
of your recovery codes.

### "Too many incorrect codes have been entered for this account, so signing in is paused."

Too many wrong codes were entered. Try again after the time the message names. If you did not enter
them, reset your password.

### "The text typed isn't this environment's key, so nothing was destroyed."

Type the environment's key exactly as the dialog shows it, in lower case.

### "Destroy works while the cloud account takes deploys"

The cloud account is frozen, being deleted or not ready, or your organization is suspended. Destroy the
environment once its account takes deploys again and its row shows **Active**.

### "Development's domain is changing"

A domain change is being prepared, or the environment moved to a new domain and hasn't deployed since.
Cancel the change, or deploy once, then destroy it.

### "The repository's destroy workflow isn't on main"

The repository's destroy workflow is missing from the environment's branch or from the repository's
default branch. Restore it there, or contact Productmind support.

### "A destroy that has started can't be stopped."

The destroy is already removing the stack. It finishes, or fails and can be tried again.

### "Productmind can't tell yet whether this destroy has started."

Productmind is still finding out from GitHub whether the destroy has started. Try again in a few
minutes.

### "The recovery code you entered is used up"

The destroy was refused after your recovery code was checked, for the reason before these words. Use
another recovery code when you try again.

### "Wait for a new code from your authenticator when you try again."

The destroy was refused after your authenticator code was checked, for the reason before these words,
and a code can't be used twice. Enter your authenticator's next code when you try again.

### A destroy failed

The row says why. A few causes:

- **The stack differs from main's newest commit or its stack settings**: its protection is lifted only
  for a stack that matches them. Cancel the destroy, deploy, then destroy again.
- **Part of the stack's protection was lifted and it still runs**: choose **Try destroy again** to
  finish, or **Cancel destroy** to put the protection back.
- **The destroy stopped partway**: choose **Try destroy again** to finish it.
- **The snapshot failed**: nothing was removed. Choose **Try destroy again**.
- **Productmind couldn't cancel the destroy**: GitHub refused a change to the repository's deploy
  settings, so a destroy already dispatched may still run. Choose **Cancel destroy** again, or detach
  the environment.

### "This repository left Productmind's hosting, so Productmind can't destroy its stack."

The application's repository is no longer hosted by Productmind. In an account Productmind manages,
delete the account to remove what it runs; in your own AWS account, remove the stack there.

## Related topics

- [Cloud accounts](/user-guide/settings/workspace-settings/cloud-accounts)
- [Deploys](/user-guide/settings/workspace-settings/deploys)
- [Stack settings](/user-guide/settings/workspace-settings/stack-settings)
- [Domains](/user-guide/settings/workspace-settings/domains)
