Default nmstate provider changes #908

Jaganathancse · 2025-03-19T09:00:06Z

This change is to update default os-net-config nmstate provider for network configuration.
jira: https://issues.redhat.com/browse/OSPRH-6809

karthiksundaravel

LGTM. Thanks

ekuris-redhat · 2025-03-19T09:29:16Z

lgtm

jpodivin

Code change looks fine, but could you please add a link to JIRA item in your PR message?

rabi · 2025-03-19T10:39:50Z

We probably have to be careful. Are all network configurations used by customers going to work with the provider switch? What's the kind of testing we've done for it. CI would test only one sample configuration.

vcandapp · 2025-03-19T10:47:41Z

We probably have to be careful. Are all network configurations used by customers going to work with the provider switch? What's the kind of testing we've done for it. CI would test only one sample configuration.

@rabi , we have an EPIC https://issues.redhat.com/browse/OSPRH-8768 which is under QE verification for now.
For FR2 atleast the plan is to enable this in only nodeset

ekuris-redhat · 2025-03-19T11:21:00Z

the PR is related to this Jira story : https://issues.redhat.com/browse/OSPRH-6809

Jaganathancse · 2025-03-19T11:41:44Z

Code change looks fine, but could you please add a link to JIRA item in your PR message?

done

softwarefactory-project-zuul · 2025-03-19T12:30:46Z

Build failed (check pipeline). Post recheck (without leading slash)
to rerun all jobs. Make sure the failure cause has been resolved before
you rerun jobs.

https://softwarefactory-project.io/zuul/t/rdoproject.org/buildset/49f28c4318cf47be931c0dc2796e3f40

✔️ openstack-k8s-operators-content-provider SUCCESS in 3h 29m 32s
❌ podified-multinode-edpm-deployment-crc RETRY_LIMIT Host unreachable in 1h 12m 02s
✔️ cifmw-crc-podified-edpm-baremetal SUCCESS in 1h 34m 44s
✔️ noop SUCCESS in 0s
❌ edpm-ansible-tempest-multinode RETRY_LIMIT Host unreachable in 1h 36m 59s
❌ adoption-standalone-to-crc-ceph-provider FAILURE in 3h 10m 54s

slagle · 2025-03-19T20:22:05Z

shouldn't there be some type of migration to stop/disable the network service if we are going to change the default? If a deployment was done with network-scripts originally, then we change the default, then the network settings need to be changed, it would then be done with nmstate. But isn't the old config put down with network-scripts still there that could conflict with the new settings?

rabi · 2025-03-20T04:56:26Z

shouldn't there be some type of migration to stop/disable the network service if we are going to change the default?

Yeah, I think for already configured nodes there won't be any change unless edpm_network_config_update is set to true during update/upgrade/subsequent deploy. But that would lead to some nodes configured with network-scripts and some with nmstate/NetworkManager. Not sure if that's something we want to get into. Also we should cleanup interfaces configured with network-scripts in https://github.com/os-net-config/os-net-config/blob/master/os_net_config/impl_nmstate.py#L2354 when provider is switched. I would suggest we don't change the default in a hurry without looking at and testing all scenarios.

Jaganathancse · 2025-03-20T05:12:18Z

shouldn't there be some type of migration to stop/disable the network service if we are going to change the default? If a deployment was done with network-scripts originally, then we change the default, then the network settings need to be changed, it would then be done with nmstate. But isn't the old config put down with network-scripts still there that could conflict with the new settings?

We are working on ifcfg provider to nmstate provider migration automation and work is inprogress.
For greenfield deployment, we are planning to make default provider is the nmstate and still we can add this parameter in the deployed env for ifcfg provider making 'fakse'.
I think this will not be blocker if any issues happened, and also we have not changed hurry. this activity is planned for and hold to avoid any issues due to this.

Jaganathancse · 2025-03-20T05:21:39Z

shouldn't there be some type of migration to stop/disable the network service if we are going to change the default?

Yeah, I think for already configured nodes there won't be any change unless edpm_network_config_update is set to true during update/upgrade/subsequent deploy. But that would lead to some nodes configured with network-scripts and some with nmstate/NetworkManager. Not sure if that's something we want to get into. Also we should cleanup interfaces configured with network-scripts in https://github.com/os-net-config/os-net-config/blob/master/os_net_config/impl_nmstate.py#L2354 when provider is switched. I would suggest we don't change the default in a hurry without looking at and testing all scenarios.

@rabi
We are suddenly planned and changing now. this activity planned for GA but hold this change to avoid issues and also had issues for NFV SRIOV and OVS DPDK deployments.
Looks like stable nmstate provider for NFV greenfield deployments, so we are planning now to make nmstate provider default.
Also this change will not be blocker in the deployment, still we can add this variable 'false' in nodeset CR to overcome deployment issues with ifcfg provider which is default now.

Also we are working on ifcfg provider to nmstate provider migration automation is in progress now.

Any how we need to update this variable 'false' for adoption ci jobs, we don't have ifcfg provider to nmstate provider migration in OSP 17 now.

Jaganathancse · 2025-03-20T05:24:03Z

shouldn't there be some type of migration to stop/disable the network service if we are going to change the default?

Yeah, I think for already configured nodes there won't be any change unless edpm_network_config_update is set to true during update/upgrade/subsequent deploy. But that would lead to some nodes configured with network-scripts and some with nmstate/NetworkManager. Not sure if that's something we want to get into. Also we should cleanup interfaces configured with network-scripts in https://github.com/os-net-config/os-net-config/blob/master/os_net_config/impl_nmstate.py#L2354 when provider is switched. I would suggest we don't change the default in a hurry without looking at and testing all scenarios.

@rabi We are suddenly planned and changing now. this activity planned for GA but hold this change to avoid issues and also had issues for NFV SRIOV and OVS DPDK deployments. Looks like stable nmstate provider for NFV greenfield deployments, so we are planning now to make nmstate provider default. Also this change will not be blocker in the deployment, still we can add this variable 'false' in nodeset CR to overcome deployment issues with ifcfg provider which is default now.

Also we are working on ifcfg provider to nmstate provider migration automation is in progress now.

Any how we need to update this variable 'false' for adoption ci jobs, we don't have ifcfg provider to nmstate provider migration in OSP 17 now.

@karthiksundaravel please add your comments

Jaganathancse · 2025-03-20T10:34:32Z

recheck

openshift-ci · 2025-06-10T04:40:21Z

New changes are detected. LGTM label has been removed.

softwarefactory-project-zuul · 2025-06-10T06:47:28Z

Build failed (check pipeline). Post recheck (without leading slash)
to rerun all jobs. Make sure the failure cause has been resolved before
you rerun jobs.

https://softwarefactory-project.io/zuul/t/rdoproject.org/buildset/ff7d0a8d6e9041a39d047a476559c871

✔️ openstack-k8s-operators-content-provider SUCCESS in 1h 57m 52s
✔️ podified-multinode-edpm-deployment-crc SUCCESS in 1h 19m 58s
❌ cifmw-crc-podified-edpm-baremetal FAILURE in 1h 38m 38s
✔️ noop SUCCESS in 0s
✔️ edpm-ansible-tempest-multinode SUCCESS in 1h 41m 52s
✔️ edpm-ansible-molecule-edpm_bootstrap SUCCESS in 5m 55s
❌ adoption-standalone-to-crc-ceph-provider RETRY_LIMIT in 43m 09s

evallesp · 2025-06-10T07:46:08Z

I see some failures cifmw-crc-podified-edpm-baremetal:
TASK [osp.edpm.edpm_libvirt : Install addditional packages name={{ edpm_libvirt_packages }}, download_only=True] *** Tuesday 10 June 2025 06:13:46 +0000 (0:00:00.125) 0:00:30.085 ********** �[1;30mFAILED - RETRYING: [edpm-compute-0]: Install addditional packages (5 retries left).�[0m �[1;30mFAILED - RETRYING: [edpm-compute-0]: Install addditional packages (4 retries left).�[0m �[1;30mFAILED - RETRYING: [edpm-compute-0]: Install addditional packages (3 retries left).�[0m �[1;30mFAILED - RETRYING: [edpm-compute-0]: Install addditional packages (2 retries left).�[0m �[1;30mFAILED - RETRYING: [edpm-compute-0]: Install addditional packages (1 retries left).�[0m �[0;31mfatal: [edpm-compute-0]: FAILED! => {"attempts": 5, "changed": false, "failures": [], "msg": "Failed to download packages: Curl error (6): Couldn't resolve host name for https://mirrors.fedoraproject.org/metalink?repo=epel-9&arch=x86_64&infra=$infra&content=centos [Could not resolve host: mirrors.fedoraproject.org]", "rc": 1, "results": []}�

Rechecking and holding the node to see what's going on

evallesp · 2025-06-10T07:46:54Z

recheck

Jaganathancse · 2025-06-10T07:52:44Z

I see some failures cifmw-crc-podified-edpm-baremetal: TASK [osp.edpm.edpm_libvirt : Install addditional packages name={{ edpm_libvirt_packages }}, download_only=True] *** Tuesday 10 June 2025 06:13:46 +0000 (0:00:00.125) 0:00:30.085 ********** �[1;30mFAILED - RETRYING: [edpm-compute-0]: Install addditional packages (5 retries left).�[0m �[1;30mFAILED - RETRYING: [edpm-compute-0]: Install addditional packages (4 retries left).�[0m �[1;30mFAILED - RETRYING: [edpm-compute-0]: Install addditional packages (3 retries left).�[0m �[1;30mFAILED - RETRYING: [edpm-compute-0]: Install addditional packages (2 retries left).�[0m �[1;30mFAILED - RETRYING: [edpm-compute-0]: Install addditional packages (1 retries left).�[0m �[0;31mfatal: [edpm-compute-0]: FAILED! => {"attempts": 5, "changed": false, "failures": [], "msg": "Failed to download packages: Curl error (6): Couldn't resolve host name for https://mirrors.fedoraproject.org/metalink?repo=epel-9&arch=x86_64&infra=$infra&content=centos [Could not resolve host: mirrors.fedoraproject.org]", "rc": 1, "results": []}�

Rechecking and holding the node to see what's going on

also adoption-standalone-to-crc-ceph-provider failing

Jaganathancse · 2025-06-10T09:53:44Z

rechek

softwarefactory-project-zuul · 2025-06-10T11:15:41Z

Build failed (check pipeline). Post recheck (without leading slash)
to rerun all jobs. Make sure the failure cause has been resolved before
you rerun jobs.

https://softwarefactory-project.io/zuul/t/rdoproject.org/buildset/d67186456c0c409fbf5980251578837c

✔️ openstack-k8s-operators-content-provider SUCCESS in 3h 27m 30s
✔️ podified-multinode-edpm-deployment-crc SUCCESS in 1h 20m 45s
❌ cifmw-crc-podified-edpm-baremetal FAILURE in 1h 45m 59s
✔️ noop SUCCESS in 0s
❌ edpm-ansible-tempest-multinode FAILURE in 1h 44m 48s
✔️ edpm-ansible-molecule-edpm_bootstrap SUCCESS in 6m 11s
✔️ adoption-standalone-to-crc-ceph-provider SUCCESS in 3h 10m 56s

Jaganathancse · 2025-06-10T11:55:18Z

recheck

softwarefactory-project-zuul · 2025-06-10T13:54:28Z

Build failed (check pipeline). Post recheck (without leading slash)
to rerun all jobs. Make sure the failure cause has been resolved before
you rerun jobs.

https://softwarefactory-project.io/zuul/t/rdoproject.org/buildset/e62674f7bc404d67b7da1645afb1050a

✔️ openstack-k8s-operators-content-provider SUCCESS in 1h 58m 01s
✔️ podified-multinode-edpm-deployment-crc SUCCESS in 1h 20m 30s
❌ cifmw-crc-podified-edpm-baremetal FAILURE in 1h 43m 39s
✔️ noop SUCCESS in 0s
✔️ edpm-ansible-tempest-multinode SUCCESS in 1h 44m 37s
✔️ edpm-ansible-molecule-edpm_bootstrap SUCCESS in 6m 25s
❌ adoption-standalone-to-crc-ceph-provider FAILURE in 46m 15s

This change is to update default os-net-config nmstate provider for network configuration and also fixing DNS config update issue when nmste provider or nmsate tool used for network configuration. And also defaulting DNS update config 'edpm_bootstrap_network_resolvconf_update' true for default nmstate provider and alse updated false for ifcfg provider based edpm_network_config_nmstte variable.

softwarefactory-project-zuul · 2025-06-10T22:47:56Z

Build failed (check pipeline). Post recheck (without leading slash)
to rerun all jobs. Make sure the failure cause has been resolved before
you rerun jobs.

https://softwarefactory-project.io/zuul/t/rdoproject.org/buildset/fa517245e07b4d3ca25cda1f09c4bf15

✔️ openstack-k8s-operators-content-provider SUCCESS in 3h 24m 37s
✔️ podified-multinode-edpm-deployment-crc SUCCESS in 1h 12m 10s
❌ cifmw-crc-podified-edpm-baremetal FAILURE in 26m 49s
✔️ noop SUCCESS in 0s
✔️ edpm-ansible-tempest-multinode SUCCESS in 1h 28m 37s
✔️ edpm-ansible-molecule-edpm_bootstrap SUCCESS in 5m 46s
✔️ adoption-standalone-to-crc-ceph-provider SUCCESS in 3h 08m 47s

Jaganathancse · 2025-06-11T03:25:48Z

recheck

Jaganathancse · 2025-06-11T05:59:00Z

roles/edpm_network_config/meta/argument_specs.yml

        type: bool
        description: "Cleanup network interfaces not in network config"
        default: true
+      edpm_bootstrap_network_resolvconf_update:


Keeping same name used in edpm_bootrap role for edpm_network_config, will propose new PR later based on edpm_network_config role and need changes in other datapalane adoption project also.

karthiksundaravel

Thanks. lgtm

openshift-ci · 2025-06-11T09:11:21Z

[APPROVALNOTIFIER] This PR is APPROVED

This pull-request has been approved by: bshephar, Jaganathancse, jpodivin, karthiksundaravel, vcandapp

The full list of commands accepted by this bot can be found here.

The pull request process is described here

Details

Needs approval from an approver in each of these files:

~~OWNERS~~ [jpodivin]

Approvers can indicate their approval by writing /approve in a comment
Approvers can cancel approval by writing /approve cancel in a comment

Default value was changed to true (NM is used instead of ifcfg) with [1]. With the new default value, os-net-config fails on BGP setups. Related-Bug: OSPRH-17472 [1] openstack-k8s-operators/edpm-ansible#908

[BGP] Set edpm_network_config_nmstate=false Default value was changed to true (NM is used instead of ifcfg) with [1]. With the new default value, os-net-config fails on BGP setups. Related-Bug: OSPRH-17472 [1] openstack-k8s-operators/edpm-ansible#908 Reviewed-by: Luca Miccini Reviewed-by: Candido Campos Rivas Reviewed-by: John Fulton <[email protected]>

Default value was changed to true (NM is used instead of ifcfg) with [1]. With the new default value, os-net-config fails on BGP setups. Related-Bug: OSPRH-17472 [1] openstack-k8s-operators/edpm-ansible#908

openshift-ci bot requested review from frenzyfriday and viroel March 19, 2025 09:00

Jaganathancse requested review from dsneddon, karthiksundaravel and vcandapp and removed request for frenzyfriday and viroel March 19, 2025 09:00

karthiksundaravel approved these changes Mar 19, 2025

View reviewed changes

openshift-ci bot assigned karthiksundaravel Mar 19, 2025

openshift-ci bot added the lgtm label Mar 19, 2025

Jaganathancse requested review from abays, bshephar and jpodivin March 19, 2025 09:21

jpodivin reviewed Mar 19, 2025

View reviewed changes

bshephar approved these changes Mar 19, 2025

View reviewed changes

openshift-ci bot assigned bshephar Mar 19, 2025

openshift-ci bot added the approved label Mar 19, 2025

Jaganathancse requested a review from jpodivin March 19, 2025 11:42

openshift-ci bot removed the lgtm label Jun 10, 2025

Jaganathancse force-pushed the default_nmstate_provider branch from a6e0023 to 5e4ecbd Compare June 10, 2025 04:48

Jaganathancse requested a review from danpawlik June 10, 2025 04:50

Jaganathancse force-pushed the default_nmstate_provider branch 4 times, most recently from 846e175 to 6a0fc76 Compare June 10, 2025 19:11

Jaganathancse force-pushed the default_nmstate_provider branch from 6a0fc76 to cf54fe3 Compare June 10, 2025 19:21

Jaganathancse commented Jun 11, 2025

View reviewed changes

eshulman2 added the lgtm label Jun 11, 2025

karthiksundaravel approved these changes Jun 11, 2025

View reviewed changes

openshift-merge-bot bot merged commit 50a12d8 into openstack-k8s-operators:main Jun 11, 2025
34 checks passed

eduolivares mentioned this pull request Jun 13, 2025

[BGP] Set edpm_network_config_nmstate=false openstack-k8s-operators/architecture#566

Merged

rabi mentioned this pull request Aug 13, 2025

Let NetworkManager manage resolv.conf after all downloads #1001

Closed

Default nmstate provider changes #908

Default nmstate provider changes #908

Uh oh!

Conversation

Jaganathancse commented Mar 19, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

karthiksundaravel left a comment

Choose a reason for hiding this comment

Uh oh!

ekuris-redhat commented Mar 19, 2025

Uh oh!

jpodivin left a comment

Choose a reason for hiding this comment

Uh oh!

rabi commented Mar 19, 2025

Uh oh!

vcandapp commented Mar 19, 2025

Uh oh!

ekuris-redhat commented Mar 19, 2025

Uh oh!

Jaganathancse commented Mar 19, 2025

Uh oh!

softwarefactory-project-zuul bot commented Mar 19, 2025

Uh oh!

slagle commented Mar 19, 2025

Uh oh!

rabi commented Mar 20, 2025

Uh oh!

Jaganathancse commented Mar 20, 2025

Uh oh!

Jaganathancse commented Mar 20, 2025

Uh oh!

Jaganathancse commented Mar 20, 2025

Uh oh!

Jaganathancse commented Mar 20, 2025

Uh oh!

openshift-ci bot commented Jun 10, 2025

Uh oh!

softwarefactory-project-zuul bot commented Jun 10, 2025

Uh oh!

evallesp commented Jun 10, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

evallesp commented Jun 10, 2025

Uh oh!

Jaganathancse commented Jun 10, 2025

Uh oh!

Jaganathancse commented Jun 10, 2025

Uh oh!

softwarefactory-project-zuul bot commented Jun 10, 2025

Uh oh!

Jaganathancse commented Jun 10, 2025

Uh oh!

softwarefactory-project-zuul bot commented Jun 10, 2025

Uh oh!

softwarefactory-project-zuul bot commented Jun 10, 2025

Uh oh!

Jaganathancse commented Jun 11, 2025

Uh oh!

Jaganathancse Jun 11, 2025

Choose a reason for hiding this comment

Uh oh!

karthiksundaravel left a comment

Choose a reason for hiding this comment

Uh oh!

openshift-ci bot commented Jun 11, 2025

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

13 participants

Jaganathancse commented Mar 19, 2025 •

edited

Loading

evallesp commented Jun 10, 2025 •

edited

Loading