Bug 1819726 - Bridge network interface only created on one node, and not another.
Summary: Bridge network interface only created on one node, and not another.
Keywords:
Status: CLOSED WORKSFORME
Alias: None
Product: Container Native Virtualization (CNV)
Classification: Red Hat
Component: Networking
Version: 2.3.0
Hardware: Unspecified
OS: Unspecified
unspecified
unspecified
Target Milestone: ---
: future
Assignee: Quique Llorente
QA Contact: Meni Yakove
URL:
Whiteboard:
Depends On:
Blocks:
TreeView+ depends on / blocked
 
Reported: 2020-04-01 11:43 UTC by Tomasz Barański
Modified: 2020-09-03 12:12 UTC (History)
3 users (show)

Fixed In Version:
Doc Type: If docs needed, set a value
Doc Text:
Clone Of:
Environment:
Last Closed: 2020-09-03 12:12:47 UTC
Target Upstream Version:
Embargoed:


Attachments (Terms of Use)
netutil daemonset spec (1.81 KB, text/plain)
2020-04-01 11:43 UTC, Tomasz Barański
no flags Details
Node network config policy (298 bytes, text/plain)
2020-04-01 11:44 UTC, Tomasz Barański
no flags Details
Longs fro nmstate-handler-worker (82.53 KB, text/plain)
2020-04-01 11:45 UTC, Tomasz Barański
no flags Details

Description Tomasz Barański 2020-04-01 11:43:27 UTC
Created attachment 1675387 [details]
netutil daemonset spec

Description of problem:

I'm running cnv-tests on a local kubevirtci cluster using k8s-1.17 provider. A lot of networking test fails due to failure creating bridge network. I was able to narrow down the problem to failure in applying the following policy:

{'apiVersion': 'nmstate.io/v1alpha1', 'kind': 'NodeNetworkConfigurationPolicy', 'metadata': {'name': 'test-network-operator'}, 'spec': {'desiredState': {'interfaces': [{'name': 'br1test', 'type': 'linux-bridge', 'state': 'up', 'bridge': {'options': {'stp': {'enabled': False}}, 'port': []}}]}}}


How reproducible:
100%

Steps to Reproduce:
1. Start kubevirt cluster (cd kubevirtci; make cluster-up; cd cnv-tests; UPSTREAM=1 make cluster-install-hco)
2. Create netutils daemonset (kubectl create -f netutil-deamon-set.yaml, attached)
3. Create namespace (kubectl create namespace cluster-addons-operator-test-network-addons-operator)
4. Create policy (kubectl create -f node-network-config.yaml, attached)

Actual results:
Interface br1test created on node01 but not on node02.

Expected results:
Interface br1test created on node01 and node02.

Additional info:

$ export | grep KUBE
declare -x KUBEVIRT_NUM_NODES="2"
declare -x KUBEVIRT_NUM_SECONDARY_NICS="1"
declare -x KUBEVIRT_PROVIDER="k8s-1.17"

Comment 1 Tomasz Barański 2020-04-01 11:44:32 UTC
Created attachment 1675388 [details]
Node network config policy

Comment 2 Tomasz Barański 2020-04-01 11:45:42 UTC
Created attachment 1675389 [details]
Longs fro nmstate-handler-worker

Comment 3 Petr Horáček 2020-04-01 11:50:05 UTC
Thanks Tomasz!

Since this is not the only issue with reporting on knmstate, we may be able to piggy-pack it to 2.3. We will need an ack for blocker though.

Comment 4 Petr Horáček 2020-04-02 10:22:42 UTC
Taking a second look. This happened only on virtualized provider and we were not able to reproduce it. I'm moving it to future and keeping this in case we reproduce it on D/S.

Comment 5 Nelly Credi 2020-04-06 06:47:32 UTC
should this move back to NEW?

Comment 6 Petr Horáček 2020-09-03 12:12:47 UTC
Please feel to reopen this if it happens again.


Note You need to log in before you can comment on or make changes to this bug.