Platform overview
Two planes, one GitOps repo. A RHEL management plane runs Symphony’s brain (EGO/SOAM, Host Factory) as a traditional install inside VMs; a Windows compute plane clones, syspreps, and joins itself from a golden image. Today both planes live on one self-hosted cluster on AWS; the IBM Cloud burst site is designed but not yet built.
The management plane (RHEL VMs, GitOps-reconciled)
Masters run the full Symphony management stack — EGO, SOAM, Host Factory — as a traditional install inside the VM. No daemon is containerized. Detail: Symphony management.
The Windows compute plane (golden image → clone → join)
Clone-to-AgentConnected measured at ~5 min on real reclones (VM Running in
~2-3 min from delete, guest agent up by ~5 min). Full clone-to-joined timing is the
Phase 2 exit measurement and is not yet closed out — see
SWOT & sign-off. Detail:
golden image and
compute pool.
Topology context: one site built, one designed
| Site | Status | What it is |
|---|---|---|
|
Built — self-hosted OpenShift 4.20, IPI |
Both planes live here today: RHEL management VMs and the Windows |
ROKS burst site (IBM Cloud us-east) |
Designed, not built (Phase 3+) |
Windows compute-host VMs only, provisioned by the custom Host Factory KubeVirt provider, VNI-routable back to the AWS masters with no NAT. |
sym-aws is self-hosted rather than ROSA per D9: ROSA HCP was blocked by an AWS
Marketplace billing link with no CLI/API workaround, so the platform pivoted to
installer-provisioned infrastructure, which needs only AWS credentials and a pull
secret. See ADR 0009 for the full story and
Architecture for the complete topology diagram, measured
numbers, and design invariants.
Continue to Architecture.