-
Bug
-
Resolution: Done-Errata
-
Normal
-
4.13, 4.12, 4.14
-
No
-
False
-
This is a clone of issue OCPBUGS-18113. The following is the description of the original issue:
—
Description of problem:
When the installer generates a CPMS, it should only add the `failureDomains` field when there is more than one failure domain. When there is only one failure domain, the fields from the failure domain, eg the zone, should be injected directly into the provider spec and the failure domain should be omitted. By doing this, we avoid having to care about failure domain injection logic for single zone clusters. Potentially avoiding bugs (such as some we have seen recently). IIRC we already did this for OpenStack, but AWS, Azure and GCP may not be affected.
Version-Release number of selected component (if applicable):
How reproducible:
Can be demonstrated on Azure on the westus region which has no AZs available. Currently the installer creates the following, which we can omit entirely: ``` failureDomains: platform: Azure azure: - zone: "" ```
Steps to Reproduce:
1. 2. 3.
Actual results:
Expected results:
Additional info:
- clones
-
-
- Closed
-
- is blocked by
-
-
- Closed
-
- links to
-
RHSA-2023:5006
OpenShift Container Platform 4.14.z security update
[OCPBUGS-18883] CPMS failure domains should be omitted when a single failure domain is present
Jinyun Ma
added a comment - joelspeed got it , thanks for the explanation!
then there is no issue now, move bug to VERIFIED.
Jinyun Ma
added a comment - joelspeed got it , thanks for the explanation!
then there is no issue now, move bug to VERIFIED. 
Joel Speed
added a comment - Hey, yes, the empty string for platform is what I mean when I say omitted. This makes the CPMS ignore the failure domain injection logic. If platform were Azure, then it would expect a list of azure zones, which we want to omit here, so this is as expected. If you can make a note on the test case that `platform: ""` is correct that would be good
Joel Speed
added a comment - Hey, yes, the empty string for platform is what I mean when I say omitted. This makes the CPMS ignore the failure domain injection logic. If platform were Azure, then it would expect a list of azure zones, which we want to omit here, so this is as expected. If you can make a note on the test case that `platform: ""` is correct that would be good Jinyun Ma
added a comment - Missed another scenario needs to verified, and I have one doubt, so move bug back to ON_QA.
Two scenarios tested:
1. Install cluster on region northcentralus without AZs, installation succeeded, and no master machine is recreated.
2. Install cluster with instance type `Standard_NP10s` that is only available in single zone (3), zone filed is directly set to machine provider spec instead of failureDomain. installation succeeded too, and no master machine is recreated.
But I found in both scenarios, failureDomain is not omitted, but have below contents:
spec: replicas: 3 selector: matchLabels: machine.openshift.io/cluster-api-cluster: jima414-2s55r machine.openshift.io/cluster-api-machine-role: master machine.openshift.io/cluster-api-machine-type: master state: Active strategy: type: RollingUpdate template: machineType: machines_v1beta1_machine_openshift_io machines_v1beta1_machine_openshift_io: failureDomains: platform: ""
I checked platform should be set to Azure in normal case, here is empty.
joelspeed could you help to check if this is expected? thanks.
Jinyun Ma
added a comment - Missed another scenario needs to verified, and I have one doubt, so move bug back to ON_QA.
Two scenarios tested:
1. Install cluster on region northcentralus without AZs, installation succeeded, and no master machine is recreated.
2. Install cluster with instance type `Standard_NP10s` that is only available in single zone (3), zone filed is directly set to machine provider spec instead of failureDomain. installation succeeded too, and no master machine is recreated.
But I found in both scenarios, failureDomain is not omitted, but have below contents:
spec:
replicas: 3
selector:
matchLabels:
machine.openshift.io/cluster-api-cluster: jima414-2s55r
machine.openshift.io/cluster-api-machine-role: master
machine.openshift.io/cluster-api-machine-type: master
state: Active
strategy:
type: RollingUpdate
template:
machineType: machines_v1beta1_machine_openshift_io
machines_v1beta1_machine_openshift_io:
failureDomains:
platform: ""
I checked platform should be set to Azure in normal case, here is empty.
joelspeed could you help to check if this is expected? thanks. Jinyun Ma
added a comment - verified on 4.14.0-0.nightly-2023-09-15-233408, and passed, move bug to VERIFIED.
Install cluster on region "northcentralus" and on Azure stack hub, both succeeded. Master machine is not recreated any more.
And master machine created by machine-api does not contain empty zone field.
Jinyun Ma
added a comment - verified on 4.14.0-0.nightly-2023-09-15-233408, and passed, move bug to VERIFIED.
Install cluster on region "northcentralus" and on Azure stack hub, both succeeded. Master machine is not recreated any more.
And master machine created by machine-api does not contain empty zone field. 
Since the problem described in this issue should be resolved in a recent advisory, it has been closed.
For information on the advisory (Important: OpenShift Container Platform 4.14.0 bug fix and security update), and where to find the updated files, follow the link below.
If the solution does not work for you, open a new bug report.
https://access.redhat.com/errata/RHSA-2023:5006