> ## Documentation Index
> Fetch the complete documentation index at: https://dpos.xdr.ooo/llms.txt
> Use this file to discover all available pages before exploring further.

# Retry & Recovery

> Resume Mirror Manager downloads safely, preserve partial files, and use snapshot recovery for destructive DP failures.

# Retry & Recovery

## Mirror Manager interrupted by SSH disconnect

Reconnect and reopen the existing installation:

```bash theme={null}
sudo ubuntu-offline-mirror mirror-manager
```

Do not reinstall merely because the terminal disconnected.

* If Menu 2 is still running, wait for the active process.
* If Menu 2 completed, continue from the first incomplete menu step.
* If configuration and the Mirror Server IP did not change, already valid steps do not need to be repeated unnecessarily.

Running Menu 4 again immediately before the DP upgrade is a useful final readiness check.

## Interrupted downloads

Run Menu 2 again. The current workflow reuses valid data and supports partial download resume.

Do not delete ACPS `.part` files just to restart; those files contain resumable download state.

## Useful Mirror Server diagnostics

```bash theme={null}
sudo ubuntu-offline-mirror mirror-manager
ls -lt /var/log/ubuntu-mirror-automation/
cat /etc/ubuntu-mirror/dp-upgrade-mirror.status
```

## DP OS-upgrade state

The Phase 1 implementation uses durable state, command journals, integrity checks, and a central state machine. It is designed to fail closed when evidence is inconsistent.

Do not delete `/opt/aelladata/os-upgrade` or invent a fresh install over orphaned evidence merely to clear an error.

For low-level Phase 1 recovery, use the diagnostics and recovery action recommended by the current release tooling. Examples in the engineering implementation include lock recovery, false reboot-required recovery, and resume-stage recovery, but operator execution should follow the release-generated procedure.

## Rollback

The project does not implement OS or DP-runtime rollback commands.

```text theme={null}
RECOVERY_METHOD=HYPERVISOR_SNAPSHOT
```

If destructive recovery cannot be completed safely, restore the approved pre-upgrade hypervisor snapshot rather than attempting an unvalidated package or runtime downgrade.