Skip to main content
Version: dev

Install and operate

This section is for a platform engineer. If you want a cluster for an evaluation, use Quickstart on Kubernetes. That page is shorter and omits the options.

Select an installation target

TargetPageStatus
Kind, for local development and testsInstall on KubernetesReady
OpenShiftInstall on OpenShiftReady
Any cluster, with Helm and OCI chartsInstall with HelmBeta

Each target needs the same decision about identity. See Authentication modes. Select SPIFFE if you install SPIRE.

Operate the platform

  • Observability covers the traces, the network diagram and the metrics.
  • Troubleshooting covers the failures that occur most often, and the recovery for each one.

Machine size

Rossoctl has many components. On a Kind cluster with one node, the platform pods alone can request almost 4 CPUs before an agent starts.

ProfileMemoryCPUsWhat operates
Recommended18 GiB6The ambient mesh, SPIRE, the console, the backend, and a build from source.
Minimum16 GiB4The core platform and the console. You must deploy each agent from an image.
Below the minimum4 or fewerThe installation usually completes. A build from source stays in the Pending state.

The installer examines the machine and gives a warning below 18 GiB or 6 CPUs. It does not stop. These values are therefore recommendations and not limits.

Subjects that have no page yet

These subjects are necessary for production use and have no page on this site:

  • An upgrade from one release to the next release.
  • High availability, and how to add capacity.
  • More than one team on one cluster: namespace isolation, quotas and token limits.
  • Backup and restoration.

Ask in Slack. If you find a correct method, a pull request for these documents is welcome.