-
Bug
-
Resolution: Done-Errata
-
Normal
-
4.12.0
-
No
-
Agent Sprint 233, Sprint 235, Sprint 236, Sprint 238
-
4
-
False
-
-
N/A
-
Release Note Not Required
Description of problem:
This may be something we want to either add a validation for or document. It was initially found at a customer site but I've also confirmed it happens with just a Compact config with no workers.
They created an agent-config.yaml with 2 worker nodes but did not set the replicas in install-config.yaml, i.e. they did not set
compute:
- hyperthreading: Enabled
name: worker
replicas: {{ num_workers }}
This resulted in an install failure as by default 3 worker replicas are created if not defined
https://github.com/openshift/installer/blob/master/pkg/types/defaults/machinepools.go#L11
See the attached console screenshot showing that the expected number of hosts doesn't match the actual.
I've also duplicated this with a compact config. We can see that the install failed as start-cluster-installation.sh is looking for 6 hosts.
[core@master-0 ~]$ sudo systemctl status start-cluster-installation.service
● start-cluster-installation.service - Service that starts cluster installation
Loaded: loaded (/etc/systemd/system/start-cluster-installation.service; enabled; vendor preset: enabled)
Active: activating (start) since Wed 2023-03-15 14:40:04 UTC; 3min 41s ago
Main PID: 3365 (start-cluster-i)
Tasks: 5 (limit: 101736)
Memory: 1.7M
CGroup: /system.slice/start-cluster-installation.service
├─3365 /bin/bash /usr/local/bin/start-cluster-installation.sh
├─5124 /bin/bash /usr/local/bin/start-cluster-installation.sh
├─5132 /bin/bash /usr/local/bin/start-cluster-installation.sh
└─5138 diff /tmp/tmp.vIq1jH9Vf2 /etc/issue.d/90_start-install.issueMar 15 14:42:54 master-0 start-cluster-installation.sh[3365]: Waiting for hosts to become ready for cluster installation...
Mar 15 14:43:04 master-0 start-cluster-installation.sh[4746]: Hosts known and ready for cluster installation (3/6)
Mar 15 14:43:04 master-0 start-cluster-installation.sh[3365]: Waiting for hosts to become ready for cluster installation...
Mar 15 14:43:15 master-0 start-cluster-installation.sh[4980]: Hosts known and ready for cluster installation (3/6)
Mar 15 14:43:15 master-0 start-cluster-installation.sh[3365]: Waiting for hosts to become ready for cluster installation...
Mar 15 14:43:25 master-0 start-cluster-installation.sh[5026]: Hosts known and ready for cluster installation (3/6)
Mar 15 14:43:25 master-0 start-cluster-installation.sh[3365]: Waiting for hosts to become ready for cluster installation...
Mar 15 14:43:35 master-0 start-cluster-installation.sh[5079]: Hosts known and ready for cluster installation (3/6)
Mar 15 14:43:35 master-0 start-cluster-installation.sh[3365]: Waiting for hosts to become ready for cluster installation...
Mar 15 14:43:45 master-0 start-cluster-installation.sh[5124]: Hosts known and ready for cluster installation (3/6)
Since the compute section in install-config.yaml is optional we can't assume that it will be there
https://github.com/openshift/installer/blob/master/pkg/types/installconfig.go#L126
Version-Release number of selected component (if applicable):
4.12
How reproducible:
Steps to Reproduce:
1. Remove the compute section from install-config.yaml 2. Do an install 3. See the failure
Actual results:
Expected results:
Additional info:
- blocks
-
-
- Closed
-
- is cloned by
-
-
- Closed
-
- links to
-
RHEA-2023:5006
rpm
[OCPBUGS-10342] Installation fails if < 3 workers defined and number of compute replicas not set
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
Great, thanks bfournie@redhat.com
Fix verified using version 4.14.0-0.nightly-2023-06-20-065807
Compact cluster with compute replicas removed and 2 workers defined
Note that this fix ONLY adds an additional warning message, it does not fail the ISO generation. If you remove the compute section (so that by default 3 computes are used) and add 2 workers in agent-config.yaml you should get the following warning when generating the ISO
level=warning msg=The number of hosts configured as workers (2) does not match the worker replicas (3)
bmanzari yes that is the expected result when using SNO. When compute replicas is not set it defaults to 3 which is why you see that failure.
I'd recommend also testing with a Compact cluster by removing the compute replicas and defining 2 worker nodes in the agent-config.yaml hosts section.
Using version 4.14.0-0.nightly-2023-06-20-065807
Removing the compute section from install-config.yaml file results in the following error
level=error msg=failed to write asset (Agent Installer ISO) to disk: cannot generate ISO image due to configuration errors
level=fatal msg=failed to fetch Agent Installer ISO: failed to load asset "Install Config": invalid install-config configuration: Compute.Replicas: Required value: Total number of Compute.Replicas must be 0 for none platform. Found 3
Prow job
https://qe-private-deck-ci.apps.ci.l2s4.p1.openshiftapps.com/view/gs/qe-private-deck/pr-logs/pull/openshift_release/40648/rehearse-40648-periodic-ci-openshift-openshift-tests-private-release-4.14-amd64-nightly-agent-baremetal-sno-ipv4-static-p1-f14/1673261389294080000
install-config.yaml file
https://gcsweb-qe-private-deck-ci.apps.ci.l2s4.p1.openshiftapps.com/gcs/qe-private-deck/pr-logs/pull/openshift_release/40648/rehearse-40648-periodic-ci-openshift-openshift-tests-private-release-4.14-amd64-nightly-agent-baremetal-sno-ipv4-static-p1-f14/1673261389294080000/artifacts/baremetal-sno-ipv4-static-p1-f14/baremetal-lab-agent-install/artifacts/install-config.yaml
is this the expected result?
Zane Bitter
added a comment - I think we have to keep this behaviour so that the install-config has the same meaning no matter which method you use to install. But we definitely need to document it well. This was still a surprise to me and I'm pretty sure this is not the first time I had learned of it 
Zane Bitter
added a comment - I think we have to keep this behaviour so that the install-config has the same meaning no matter which method you use to install. But we definitely need to document it well. This was still a surprise to me and I'm pretty sure this is not the first time I had learned of it
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