SDN-4930: Downstream Merge [01-08-2025] #2412
Conversation
Handle host-network pods as default network. Don't return per-pod errors on startup. Remove nadController from UDNHostIsolationManager as we don't use it anymore to find pod's UDN based on NADs that exist in the namespace. Signed-off-by: Nadia Pinaeva <n.m.pinaeva@gmail.com>
Signed-off-by: Jaime Caamaño Ruiz <jcaamano@redhat.com>
udn host isolation: fix initialSync.
…face Signed-off-by: Martin Kennelly <mkennell@redhat.com>
This code isnt being used anymore. We dont expect users to upgrade directly from code which contained the legacy LRPs, therefore its safe to remove. Signed-off-by: Martin Kennelly <mkennell@redhat.com>
Signed-off-by: Martin Kennelly <mkennell@redhat.com>
L2 UDN: EgressIP hosted by primary interface (`breth0`)
If EncapIP is configured, it means it is different from the node's primary address. Do not update EncapIP when node's primary address changes. Signed-off-by: Yun Zhou <yunz@nvidia.com>
Assign network ID from network manager running in cluster manager. The network ID is included in NetInfo and annotated on the NAD along with the network name. Network managers running in zone & node controllers will read the network ID from the annotation to set it on NetInfo. On startup, network manager running in cluster manager will read the network IDs annotated on the nodes to cover for the upgrade scenario. Network IDs will still be annotated on the nodes because this PR does not transition all the code to use the network ID from the NetInfo instead of the node annotation. That will have to be done progressively. This have several benefits, among them: - NetworkID is available sooner overall since we dont have to wait for all the nodes to be annotated - No need to unmarshall the node annotation to get the network IDs, they are available in NetInfo - No need to unmashall the NAD to get the network name, can be accessed directly from the annotation. If a network is replaced with a different one with the same name, the network ID is reused as the respective network controller will not start as the previous one is stopped and cleaned up so it shouldn't be a problem. Signed-off-by: Jaime Caamaño Ruiz <jcaamano@redhat.com>
Instead of considering managed VRFs those that follow the mp<id>-udn-vrf naming template, use the table number: those vrfs associated to a table within our reserved block of table numbers are managed by us. The block right now is anything higher than RoutingTableIDStart (1000). This allows to manage VRFs with any name which is desirable if the name is going to be exposed through BGP. Signed-off-by: Jaime Caamaño Ruiz <jcaamano@redhat.com>
Anticipating that these VRF names are going to be exposed through BGP, we should to use friendlier names for our VRFs. The most natural name to use is the network name. Thus giving a cluster UDN a name below 15 characters that matches an already existing VRF not managed by ovn-k will fail. This is considered an admin problem and not an ovn-k problem for now. Signed-off-by: Jaime Caamaño Ruiz <jcaamano@redhat.com>
Signed-off-by: Jaime Caamaño Ruiz <jcaamano@redhat.com>
Signed-off-by: Jaime Caamaño Ruiz <jcaamano@redhat.com>
Signed-off-by: Jaime Caamaño Ruiz <jcaamano@redhat.com>
Was causing deadlocks in unit tests Signed-off-by: Jaime Caamaño Ruiz <jcaamano@redhat.com>
…heir subcontrollers Signed-off-by: Jaime Caamaño Ruiz <jcaamano@redhat.com>
Signed-off-by: Jaime Caamaño Ruiz <jcaamano@redhat.com>
Assuming that there is three types of controllers, being: network agnostic, network aware and network specific; we were already notifying network specific controllers of network changes. But network aware controllers, controllers for which we have a single instance capable of managing multiple networks, had no code path to be informed of netwokr changes. This commit adds a code path for that and makes the RouteAdvertisments controller aware of network changes. Changed ClusterManager to be the controller manager for cluster manager instead of secondaryNetworkClusterManager. It just makes more sense that way sice ClusterManager is the top level manager. Signed-off-by: Jaime Caamaño Ruiz <jcaamano@redhat.com>
Signed-off-by: Jaime Caamaño Ruiz <jcaamano@redhat.com>
…twork exist test Signed-off-by: Or Mergi <ormergi@redhat.com>
On CUDN cleanup is inconsistent as we see some flaky tests due to CUDN "already exist" errors, implying object are not actually deleted. Wait for CUDN object be gone when deleted Signed-off-by: Or Mergi <ormergi@redhat.com>
CUDN is cluster-scoped object, in case tests running in parallel, having random names avoids conflicting with other tests. Use random metadata.name for CUDN objects. The "isolates overlapping CIDRs" tests create objects based on the 'red' and 'blue' variables, including CUDN objects. Change the tests CUDN creation use random names and update the given 'networkAttachmentConfigParams' with the new generated name. Update 'red' & 'blue' vaiables with the generated name, carried by 'networkAttachmentConfigParams' (netConfig.name). The pod2Egress tests asserts on the CUDN object name given by 'userDefinedNetworkName'. In practice the tests netConfigParam.name is userDefinedNetworkName. Change the assertion to check the given netConfigParam. Signed-off-by: Or Mergi <ormergi@redhat.com>
Signed-off-by: nithyar <nithyar@nvidia.com>
Signed-off-by: nithyar <nithyar@nvidia.com>
e2e, CUDN: Improve stability
Reconcile RouteAdvertisements in cluster manager
Add missing enum validation for RouteAdvertisements
The NetPol test checks assigned pod IP only against IPv4 subnet which would fail on IPv6 only cluster. This commit fixes it by checking on all valid CIDRs. Signed-off-by: Periyasamy Palanisamy <pepalani@redhat.com>
The variable ginkgo_focus is misspelled as gingko_focus. As the latter var is not used anywhere else in this repo and is used to concatenate the var ginkgo_focus in the next line to ginkgoargs it seems to be a typo. Fixes: #4942 Signed-off-by: Felix Schumacher <felix.schumacher@internetallee.de>
This commit adds a new controller to import BGP learnt routes into OVN. The controller runs in ovnkube-controller so it only supports IC architecture where ovnkube-controller has kernel access on each node. Networks should register to this controller to have routes imported for them. Routes are imported into the network's gateway router. Multipath routes are supported. The controller subscribes for netlink route events. When a route is updated, the corresponding network is queued to be sync'ed. A network is also sync'ed when registered to the controller. Synchronizations are delayed by a small amount of time to prevent a series of consecutive route updates so synchornize the same network twice. Synchronizations apply the difference between current and desired state. The controller subscribes to netlink link events to learn the routing table associated to a network vrf. The network is inferred from the vrf device name. When learning the routing table, the corresponding network is queued to be sync'ed. Signed-off-by: Jaime Caamaño Ruiz <jcaamano@redhat.com>
Signed-off-by: Jaime Caamaño Ruiz <jcaamano@redhat.com>
Signed-off-by: Periyasamy Palanisamy <pepalani@redhat.com>
Signed-off-by: Periyasamy Palanisamy <pepalani@redhat.com>
Fix subnet check for assigned pod IPs
go-controller: fix typo in test script
Import learnt BGP routes into OVN
openshift-ci-robot
added
the
jira/valid-reference
|
@jluhrsen: This pull request references SDN-4930 which is a valid jira issue. Warning: The referenced jira issue has an invalid target version for the target branch this PR targets: expected the epic to target either version "4.19." or "openshift-4.19.", but it targets "openshift-4.18" instead. In response to this:
Instructions for interacting with me using PR comments are available here. If you have questions or suggestions related to my behavior, please file an issue against the openshift-eng/jira-lifecycle-plugin repository. |
openshift-merge-robot
added
the
needs-rebase
|
[APPROVALNOTIFIER] This PR is NOT APPROVED This pull-request has been approved by: jluhrsen The full list of commands accepted by this bot can be found here.
Needs approval from an approver in each of these files:
Approvers can indicate their approval by writing |
jluhrsen
force-pushed
the
d/s-merge-01-08-2025
branch
from
1759248 to
65a3e28
Compare
openshift-merge-robot
removed
the
needs-rebase
|
/test e2e-metal-ipi-ovn-ipv6-techpreview |
|
@jluhrsen: trigger 4 job(s) of type blocking for the ci release of OCP 4.19
See details on https://pr-payload-tests.ci.openshift.org/runs/ci/59d94350-ce24-11ef-9fe7-2791c9a6bd2e-0 trigger 14 job(s) of type blocking for the nightly release of OCP 4.19
See details on https://pr-payload-tests.ci.openshift.org/runs/ci/59d94350-ce24-11ef-9fe7-2791c9a6bd2e-1 |
|
/test e2e-azure-ovn-techpreview |
|
/test e2e-aws-ovn-serial |
|
/payload-aggregate periodic-ci-openshift-microshift-release-4.19-periodics-e2e-aws-ovn-ocp-conformance 10 |
|
@jluhrsen: trigger 3 job(s) for the /payload-(with-prs|job|aggregate|job-with-prs|aggregate-with-prs) command
See details on https://pr-payload-tests.ci.openshift.org/runs/ci/12975570-cea8-11ef-820f-aade8edf1c59-0 |
|
@jluhrsen: The following tests failed, say
Full PR test history. Your PR dashboard. Instructions for interacting with me using PR comments are available here. If you have questions or suggestions related to my behavior, please file an issue against the kubernetes-sigs/prow repository. I understand the commands that are listed here. |
|
/test e2e-azure-ovn-upgrade |
|
/payload-job periodic-ci-openshift-release-master-nightly-4.19-e2e-rosa-sts-ovn 1 |
|
@jluhrsen: trigger 1 job(s) for the /payload-(with-prs|job|aggregate|job-with-prs|aggregate-with-prs) command
See details on https://pr-payload-tests.ci.openshift.org/runs/ci/417d8390-d1e7-11ef-8bd8-742fa6af9ebf-0 |
|
/payload-job periodic-ci-openshift-release-master-ci-4.19-upgrade-from-stable-4.18-e2e-aws-ovn-upgrade 10 |
|
@jluhrsen: trigger 1 job(s) for the /payload-(with-prs|job|aggregate|job-with-prs|aggregate-with-prs) command
See details on https://pr-payload-tests.ci.openshift.org/runs/ci/78df37c0-d1e7-11ef-945e-d8e140fc0948-0 |
|
/payload 4.19 ci blocking |
|
@jluhrsen: trigger 4 job(s) of type blocking for the ci release of OCP 4.19
See details on https://pr-payload-tests.ci.openshift.org/runs/ci/a3a513d0-d1e7-11ef-97fe-20124f3e1665-0 |
📑 Description
Fixes #
Additional Information for reviewers
✅ Checks
How to verify it