If a gateway is deleted (e.g. during a downscaling event), remote clusters would continue to attempt to send traffic to the gateway IP address until they received an update that the gateway’s IP address was removed. This may cause
503 errors for HTTP traffic or
000 errors for passthrough cross cluster traffic.
Since TSB 1.6, you can delay gateway deletions by a configurable period to provide sufficient time for the gateway’s IP address removal to propagate across other clusters to avoid
000 errors. Currently this feature is disabled by default.
Enable Gateway Deletion Hold Webhook
In order to enable a gateway deletion hold webhook in your control plane, you will need to
xcp component in
ControlPlane CR or Helm values and add the following environment variables:
ENABLE_GATEWAY_DELETE_HOLDwith value set to
GATEWAY_DELETE_HOLD_SECONDS. This is optional with default is 10 seconds
- name: ENABLE_GATEWAY_DELETE_HOLD
- name: GATEWAY_DELETE_HOLD_SECONDS
This will delay gateway deletions for 20 seconds while the deleted gateway IP is removed from remote clusters.