16.2.x to 17.0.0

Version 17.0.0 changes how test results are stored. Three things happen when you upgrade, and one of them needs an action on your side.

Back up your database before upgrading. As with any major version, the upgrade cannot be reverted.

Your load generators must be upgraded with the server

A 16.2 load generator cannot report to a 17.0 server. Pull the load generator images again on every machine that runs an on-premise agent, so they run the 17.0.0 tag like the rest of the stack, and restart the agents. If you launch load generators on Kubernetes, update the image tags in your own configuration too.

Leaving an agent behind does not corrupt anything - the test simply reports nothing.

The definitions are upgraded automatically at startup

Reports, report templates, monitoring connections and SLA profiles are rewritten to the new format when the server starts, with no action from you. The server is unavailable while it runs.

The upgrade also repairs SLA thresholds that measured the whole scenario instead of the action they were set on. If you edit an SLA profile afterwards, its thresholds keep the action they measure.

Test results are updated one by one, on demand

Your existing test results are not converted at startup: a history of test results can be very large, and converting all of it would keep the platform down for a long time. Instead, each result is updated the first time someone opens its report - see Updating report data.

Nothing is deleted, and a result that has never been reopened stays exactly as it is.

If you call the REST API

The analysis endpoints changed in a breaking way. Every change and its replacement is listed in REST API changes: 16.2 to 17.0. Update your scripts and pipelines before or right after the upgrade.