GeezerVison

POST, PUT, and PATCH /v0/pipelines now refuse a runtime config with memory above the runner's maximum, or with storage autoscaling when the runner does not enable it, with a 400 InvalidRuntimeConfig. A PATCH of deployment_config that resizes a running pipeline refuses memory above the maximum, with a 400 InvalidDeploymentPatch. Before, a resize the runner refused returned 200, went only to the pipeline log, and failed the next start.

With #7712 (storage size and class allowlists) and #7656 (the resize PATCH checks them), every field that can change while a pipeline runs is checked against the runner's limits. CPU has no runner limit.

• Behavior change on all installations, in the changelog: memory_min above memory_max is refused, as storage_min above storage_max already is. A memory refusal reads "unsupported memory: ...".
• Otherwise each new flag is unset by default, and an unset flag is not checked, so the open-source manager and a local install see no other change.
• Only new values are checked: a stored runtime config still loads.
• The "Storage limits" docs section becomes "Resource limits", with both rows.

Settings that cannot change while a pipeline runs, such as resources.namespace, service_account_name, or env, are left to the runner, which refuses them at start with a message that names the Helm value.

The Helm side is feldera/cloud#2165, which must merge after this.

Testing: unit tests for each refusal and for parsing the flags as the chart passes them, and endpoint tests for POST, PUT, PATCH of runtime_config (including the deprecated memory_mb_max and an installation without limits), and PATCH of deployment_config. Each check was removed in turn, and a test failed. Platform tests were not run.