---
id: platform/recovery-etcd
title: "Recovery drill: etcd"
description: Extract Bex CRs from an etcd snapshot into a throwaway local etcd, then optionally apply reviewed manifests to a restore-* context.
keywords: [bex, recovery, etcd, snapshot]
last_updated: 2026-09-23
---

# Recovery drill: etcd

Use `scripts/restore-etcd.sh` from the Bex repository (pinned revision in
[Platform recovery rehearsals](./recovery.md)). The script restores a snapshot
into a **local throwaway etcd**, extracts selected CR prefixes, and can apply
reviewed JSON only to a kube context whose name starts with `restore-`. It cannot
perform in-place / Path B recovery against the live control plane.

## Prepare

- Confirm a decryptable etcd snapshot object exists (or pass
  `--snapshot s3://…`).
- Install the etcd image pin you will use (`--image`).
- Choose an empty `--output-dir` for extracted manifests.

## DRY_RUN

```bash
cd /path/to/bex
DRY_RUN=1 scripts/restore-etcd.sh --snapshot latest --output-dir /tmp/etcd-drill
```

Review planned S3/etcd steps. Expect object-store access when resolving
`latest`.

## Extract

```bash
scripts/restore-etcd.sh --snapshot latest --output-dir /tmp/etcd-drill
```

Default prefixes cover Apps, Databases, and KeyValues under
`/registry/app.bex.co/…`. Add `--prefix` for additional registry paths.
Extraction does not contact Kubernetes.

## Optional apply

After human review of the JSON files:

```bash
scripts/restore-etcd.sh --apply-dir /tmp/etcd-drill \
  --target-context restore-drill \
  --confirm APPLY-restore-drill
```

Only the reviewed files are applied, and only to a `restore-*` context.

## Verify and clean up

Confirm the expected CRs exist in the throwaway context, then delete that
context/cluster per your local lab practice. Do not treat apply success as a
live routing restore.