> For the complete documentation index, see [llms.txt](https://wiki.datadike.com/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://wiki.datadike.com/pam/deploy-operate/upgrade.md).

# Upgrade

What is established about upgrading the solution, the preparation you own before one, the version-specific traps on record, and how to verify an upgrade afterwards.

{% hint style="warning" %}
**The upgrade procedure must be validated against the two versions involved — the one installed and the one you are moving to.** No single sequence of steps applies to every pair of versions: at least one release requires a restart, and one range of releases dropped a capability that a later release restored. Establish both version numbers first, read the release notes for everything in between, and confirm the procedure with DataDike implementation support before you start.
{% endhint %}

## What this is and when to use it

This page covers what happens around an upgrade: the preparation that is yours to do, the version-specific facts on record, and the checks that tell you the upgrade landed.

It does not give an upgrade command sequence, because there is none to give honestly.

{% hint style="danger" %}
**The commands, the order of operations and the rollback path come from DataDike implementation support or your reseller**, for your specific installation and version pair. This page describes the upgrade *model* and records version-specific changes. Do not assemble a procedure from fragments — an upgrade run from a guessed sequence is a data-loss event, not a failed change.
{% endhint %}

## What is established about upgrades

| Fact                                                                                                                                                                                                                                               | Consequence for you                                                                                                                                                                      |
| -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| In the SaaS model, hosting, infrastructure and operation are DataDike's responsibility, and subscription licensing includes continuous version updates — fixes and new features — at no additional cost for the term, with vendor support 24x7x365 | You do not run the upgrade. Confirm the window and read the release notes                                                                                                                |
| Components can be updated individually, without stopping the product                                                                                                                                                                               | Stated as a capability of the architecture, not as a procedure. It does not mean any given upgrade is non-disruptive — the release notes are what tell you whether a restart is involved |
| On-premises installations run active/active                                                                                                                                                                                                        | A node can be out of service during the change without the service stopping, but only if the change is actually applied node by node                                                     |
| Operating system fixes and updates are published and signed by DataDike and applied in a controlled way, with platform compatibility preserved at each update and a managed lifecycle that avoids end-of-life components                           | The OS and the application are updated through the vendor's channel, not from upstream distribution repositories                                                                         |
| The appliance console offers **Reboot** and **Shutdown**                                                                                                                                                                                           | Use these rather than improvising a shutdown of the appliance                                                                                                                            |

## Before you start

Everything below is preparation you own, whichever party executes the upgrade.

1. **Write down the installed version and the target version.** Every fact on this page and in the release notes is conditional on those two numbers. Without both, nothing here can be applied.
2. **Read the release notes for every version between them**, not just the target's. List each item that requires a restart, changes a default, or removes or restores a capability. See [Release notes](https://wiki.datadike.com/pam/release-notes.md).
3. **Take a backup and restore it somewhere.** An untested backup is not a rollback plan. See [Back up and restore](https://wiki.datadike.com/pam/deploy-operate/backup-and-restore.md).
4. **Confirm license validity for the window.** All functions stay disabled until a license is imported, and an expired license leaves a seven-day grace period in which to import the new one. An upgrade window is a poor time to discover the license lapsed.
5. **Schedule outside the nightly cleanup.** The maintenance and cleanup tasks run in the early hours — the source cites 02:00 for the daily cleanup jobs and recommends keeping system updates and worker restarts out of the 02:00–04:00 window so that automated tasks are not interrupted mid-run.
6. **Agree who executes the change**, and open the request with support stating both version numbers.
7. **Record what a healthy system looks like now**, before the change, so that "was it like that before?" has an answer afterwards: the current `/api/health/` response, the component list with every status, and both task workers Online. See [Monitoring and health](https://wiki.datadike.com/pam/deploy-operate/monitoring.md).

## Version-specific facts on record

These come from the release notes and are the reason the warning at the top of this page exists. They apply only to the versions named.

| Version or range       | What it changes                                                                                                                                                                                                                | What it means for you                                                                                         |
| ---------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ | ------------------------------------------------------------------------------------------------------------- |
| 5.0.1                  | Enforces TLS 1.3 communication between the Gateway VM and the server                                                                                                                                                           | The release note states a restart is required                                                                 |
| 5.0 through 5.0.9      | Editing an asset to give it a custom friendly name was removed in error                                                                                                                                                        | If you are on this range, the capability is missing — this is a known regression, not a configuration problem |
| 5.1.0 (18 August 2026) | Friendly-name editing reintroduced. Environments that already had individual renames do not lose the customisation                                                                                                             | The fix for the range above                                                                                   |
| 04 May 2026 note       | Reliability and stability improvements to the upgrade process of the gateway VM, with reduced downtime between versions. The note's heading says Broker VM while its body says Gateway VM — treat the component as unconfirmed | Broker and gateway VMs follow their own upgrade path, separate from the platform's                            |

{% hint style="warning" %}
**A capability missing after an upgrade is not always a misconfiguration.** The 5.0–5.0.9 friendly-name regression is the documented case: the setting was gone from the console, and no amount of permission or policy work would bring it back. Before troubleshooting a missing feature, check whether the release notes record it as removed.
{% endhint %}

## Steps

The execution belongs to support. The change around it is yours, and it has a shape:

1. Establish the installed and target versions.
2. Read every release note between them and write down the items that require a restart or change behaviour.
3. Back up, and prove the backup by restoring it into a test instance.
4. Capture the pre-change health baseline described under **Before you start**.
5. Open the change with support, stating both versions and the items you found in step 2.
6. Agree a window outside the nightly cleanup hours.
7. Run the upgrade with support.
8. Verify, using the checks below, before declaring the window closed.

Do not compress steps 3 and 4. They are the two that make the difference between a rollback and an incident.

## How to verify it worked

An upgrade that returns a login page has proved almost nothing. Work outward from the dependencies:

1. **`/api/health/`** returns `status`, `db_status` and `redis_status` all true. On a cluster, poll the virtual address **and each node individually** — a node left on the old version, or left out of the restart, is invisible behind the virtual address.
2. **System Settings › Components** shows every component **Normal** and none **Offline**. Compare against the baseline you captured: a component that is present but was Normal before and is High load now is a finding.
3. **The task engine is running.** From **System Settings › Operations**, open **Monitoring**: both workers — maintenance and automation — Online, and the queues without a backlog of pending **Messages**. Then check on the Operations list that a short-cycle task's **Date last publish** advances after the upgrade. A worker that came back but is not consuming looks identical to a healthy one until you check this.
4. **One real session per protocol you actually use** — terminal, graphical, database — opened end to end, not just launched. Confirm each appears in the audit trail and that its recording plays back. Protocol connectors are separate components; one of them failing to come back is the most common partial-upgrade symptom.
5. **One automation completes.** Let a credential rotation or a connectivity test run, and confirm success in its execution history. This exercises the automation worker and the assets together.
6. **Log in through your identity provider, with MFA.** Local login working proves less than it seems.
7. **Confirm the server date and time.** If the clock is wrong, OTP codes are rejected and MFA login fails — the source names an incorrect server date and time as the cause of the invalid OTP error. A reboot during an upgrade is exactly when a clock drifts.
8. **Open the console by its HTTPS name** and confirm the certificate still serves correctly. See [Certificates and HTTPS](https://wiki.datadike.com/pam/deploy-operate/certificates.md).

## Common errors

| Symptom                                                             | Cause                                                                                                                            | Fix                                                                                            |
| ------------------------------------------------------------------- | -------------------------------------------------------------------------------------------------------------------------------- | ---------------------------------------------------------------------------------------------- |
| A capability that worked before is gone                             | Known regression in the version you moved to — the documented case is custom asset friendly names, removed between 5.0 and 5.0.9 | Check the release notes for the range before troubleshooting. For that case, 5.1.0 restores it |
| MFA or OTP codes rejected immediately after the upgrade             | Server date and time incorrect after the restart                                                                                 | Correct the server clock; the source documents this as the cause of the invalid OTP error      |
| One node in a cluster behaves differently from the others           | Not every node was updated or restarted                                                                                          | Poll `/api/health/` on each node as well as on the virtual address, and compare                |
| Every automatic task stopped after the upgrade                      | A task worker did not come back                                                                                                  | **Operations › Monitoring**, Workers tab; the task service on that server needs restarting     |
| A setting changed in the appliance configuration file has no effect | Configuration held in the appliance's `config.txt` is read at service start                                                      | The services have to be restarted after the file changes                                       |
| Gateway connections fail after moving to 5.0.1                      | That release enforces TLS 1.3 between the Gateway VM and the server, and states a restart is required                            | Restart as the release note requires; confirm the gateway side supports TLS 1.3                |
| Sessions open but nothing is recorded                               | The recording or video-processing component is offline, or the storage backend is unreachable                                    | Check **Components**, and the **Storage — Connectivity** notification                          |

## Related pages

* [Release notes](https://wiki.datadike.com/pam/release-notes.md)
* [Back up and restore](https://wiki.datadike.com/pam/deploy-operate/backup-and-restore.md)
* [Monitoring and health](https://wiki.datadike.com/pam/deploy-operate/monitoring.md)
* [Certificates and HTTPS](https://wiki.datadike.com/pam/deploy-operate/certificates.md)
