Bug 1918558

Description of problem:

When metal3 deploys Supermicro nodes with Redfish and UEFI the deployment goes fine - the nodes run inspection via fast-track, the RHCOS image is installed on the node, and the nodes boot to that image on disk. On the subsequent reboot however the node attempts to PXE boot instead of booting to disk and it installs the inspector image and eventually stays looping in IPA.  The deployment fails.

The reason appears to be because the Redfish request to set to boot to disk Persistent (BootSourceOverrideEnabled': 'Continuous):
2021-01-20 23:54:49.103 1 DEBUG sushy.connector [req-d3cc400a-5752-400d-b77d-f8b2b672de4c - - - - -] HTTP request: PATCH https://mgmt-f06-h15-000-1029p.rdu2.scalelab.redhat.com/redfish/v1/Systems/1; headers: {'OData-Version': '4.0'}; body: {'Boot': {'BootSourceOverrideTarget': 'Hdd', 'BootSourceOverrideEnabled': 'Continuous'}}


Is then followed a request to boot to disk WITHOUT the BootSourceOverrideEnabled set:
2021-01-20 23:54:52.045 1 DEBUG sushy.connector [req-d3cc400a-5752-400d-b77d-f8b2b672de4c - - - - -] HTTP request: PATCH https://mgmt-f06-h15-000-1029p.rdu2.scalelab.redhat.com/redfish/v1/Systems/1; headers: {'OData-Version': '4.0'}; body: {'Boot': {'BootSourceOverrideTarget': 'Hdd'}}

This is then followed by a set of boot mode (to fix https://bugzilla.redhat.com/show_bug.cgi?id=1888072):
2021-01-20 23:54:53.360 1 DEBUG sushy.connector [req-d3cc400a-5752-400d-b77d-f8b2b672de4c - - - - -] HTTP request: PATCH https://mgmt-f06-h15-000-1029p.rdu2.scalelab.redhat.com/redfish/v1/Systems/1; headers: {'OData-Version': '4.0'}; body: {'Boot': {'BootSourceOverrideMode': 'UEFI'}}

As a result, before the boot to disk the Redfish settings are:
{
  "BootSourceOverrideEnabled": "Once",
  "BootSourceOverrideMode": "UEFI",
  "BootSourceOverrideTarget": "Hdd",

This works fine for the first boot but after that the Redfish settings read are:
  "BootSourceOverrideEnabled": "Disabled",
  "BootSourceOverrideMode": "Legacy",
  "BootSourceOverrideTarget": "None",

So all subsequent boots will be to PXE which is the default.  Note that we tried to change the Boot Order in BIOS so that HDD is first.  It would seem that if the BootSourceOverrideTarget was not set then it would use this boot order but it didn't work.


Version-Release number of selected component (if applicable):

This is seen on 4.7 build from 12/18

# rpm -qa | grep ironic
python3-ironic-prometheus-exporter-0.0.1-0.20190712090404.f7e9344.el8ost.noarch
python3-ironic-lib-4.4.1-0.20201120173800.e1b5e12.el8.noarch
openstack-ironic-common-16.0.3-0.20201219231205.4ae5375.el8.noarch
openstack-ironic-conductor-16.0.3-0.20201219231205.4ae5375.el8.noarch
openstack-ironic-api-16.0.3-0.20201219231205.4ae5375.el8.noarch

How reproducible:

Happens every time.

Steps to Reproduce:
1. Run ansible install script to deploy the cluster.

Actual results:  Nodes boot to PXE after first booting to disk after image installed.


Expected results: Nodes boot to disk every time after image is installed.


Additional info:
Its likely the issue is here in sushy - 
https://github.com/openstack/ironic/blob/master/ironic/drivers/modules/redfish/management.py#L96

The BootSourceOverrideTarget won't be set if its getting set to the same value.  However Supermicro requires it to be set each time.