# Migration Guide (2026)

The SciServer platform is being migrated from its original bare-metal
infrastructure to a cloud-based setup. This transition addresses hardware
limitations, improves scalability, and modernizes the platform. While we’ve
preserved your data and workflows, some changes require adjustments on your
part.

**Migration Window**: The migration will occur during the week of
**April 20th–24th, 2026**. A full week of downtime is scheduled to ensure a
smooth transition.

This guide explains what’s changed, how to prepare, and what to do after the
migration.

---

## Pre-Migration Checklist

- **Save Python environments** (`pip freeze > requirements.txt` or
  `conda env export > environment.yml`).
- **Coordinate with collaborators** to re-share folders
  *immediately after migration*.
- **Test workflows** in new container images if you rely on older versions
  (pre-2024).

Consider reviweing the relevant recommendations under
[Compute Best Practices](./best_practices/persistent_environments.md)

---

## What’s Changed

### Accounts and Passwords

- **Usernames and email accounts** remain unchanged.
- **Password reset required**: You must reset your password to log in after the
  migration.
  - Check your email for a reset link or request one via the
    [login page](https://sciserver.mpe.mpg.de/login-portal/reset-password).

---

### Shared Folders (⚠️ Critical)

- **Shared folders will NOT be re-shared automatically**. Folder owners must
  re-share them **manually** after migration.
  - If you depend on shared data, remind owners to re-share access.
  - If you’re missing access post-migration, contact the folder owner directly.

---

### Container Images

- Older container images (pre-2024) are **deprecated**.
- Test your workflows in new container images *before* the migration to identify
  compatibility issues.
- If you encounter problems, reach out via the [Support & Contact](./support.md)
  page.

---

### Files and Sharing

- **All files were copied** to the new infrastructure.
- Verify file access post-migration, including shared folders.
- If you’re missing files or access, contact the folder owner or support.

---

### Compute Containers

- **Performance changes**: Compute containers now run on a swarm of smaller
  servers.
  - **Fairer resource distribution**: No single user can monopolize resources.
  - **Better stability**: Previous infrastructure often suffered from degraded
    performance when a couple users had heavy workloads.
- **Workarounds for heavy tasks**:
  - Use [compute jobs](./compute_jobs.md) for resource-intensive workflows.
  - Explore the
    [MPCDF Viper supercomputer](https://www.mpcdf.mpg.de/services/supercomputing/viper)
    for demanding workloads.

---

## Post-Migration Steps

1. **Reset your password** and log in.
2. **Verify file access**, including shared folders.
3. **Recreate Python environments** using saved dependency lists.
4. **Test compute containers** and adjust workflows if needed.

---

## FAQ

### Why do I need to reset my password?

The migration required updates to our authentication system. Resetting your
password ensures secure access to the new infrastructure.

### What happened to my shared folders?

Shared folders were not automatically re-shared. Owners must re-share them
manually. If you’re missing files, contact the folder owner.

### Why do my compute containers feel slower?

The new infrastructure uses a swarm of smaller servers, which may reduce
performance for individual containers. For resource-intensive tasks, use compute
jobs or the MPCDF Viper supercomputer.

### How can I save/restore my Python environment?

Use `pip freeze > requirements.txt` or `conda env export > environment.yml` to
save dependencies. Recreate environments in the new setup using these files.

### Where can I get help if something isn’t working?

Visit the [Support & Contact](./support.md) page for assistance. Use the public
Matrix channel for interactive support.