Known Limitations
This section describes issues and limitations present in Juniper Cloud-Native Router release 26.2.
Known Issues and Limitations in Juniper Cloud-Native Router Release 26.2
- JCNR-12843: Some CoS show commands from cRPD return errors or no output—When
Cloud-Native Router is deployed with a CoS configuration, the
show class-of-service scheduler-mapcommand returns a traceback error and theshow class-of-service drop-profileandshow class-of-service interfacecommands return no output. - JCNR-13009: ARP packets over IRB interface are dropped after cRPD routing
restart—When Cloud-Native Router is deployed with an IRB interface configuration, ARP
packets are dropped after you run
restart routingon cRPD. This occurs because cRPD replays the interface configuration before the VRF is fully restored, leaving the interface VRF in an invalid state and causing bridge domain derivation to fail. As a workaround, remove the physical interfaces from the L2 VRF and add them back. -
JCNR-13369: Stale sub-interfaces remain in the vRouter after a configlet add/delete operation—When you apply a configlet with multiple sub-interfaces and then delete it, some sub-interfaces remain in the vRouter and are not cleaned up. This occurs because cRPD sends add and change requests for the affected sub-interfaces after the delete operation, causing a synchronization mismatch between cRPD and the vRouter.
-
JCNR-13555: Maximum MTU supported in GCP with gVNIC is 1500 bytes—When Cloud-Native Router is deployed in GCP with gVNIC in GQI mode, the maximum supported MTU is 1500 bytes. Additionally, gVNIC requires multi-queue support on the VM instance, with a minimum of 4 queues and a maximum of 8 queues per interface. Configure the queue count when creating the VM instance. For example:
gcloud compute instances create INSTANCE_NAME \ --zone=ZONE \ --machine-type=MACHINE_TYPE \ --network-performance-configs=total-egress-bandwidth-tier=TIER_1 \ --network-interface=network=NETWORK_NAME_1,subnet=SUBNET_1,nic-type=GVNIC,queue-count=4 \ --network-interface=network=NETWORK_NAME_2,subnet=SUBNET_2,nic-type=GVNIC,queue-count=4 -
JCNR-13465: Core files are not generated for the vRouter DPDK container when Cloud-Native Router is deployed with guaranteed pods on Wind River Cloud Platform—When Cloud-Native Router is deployed using guaranteed pods on WRCP, core files are not generated for the vRouter DPDK container even when the
starlingx.io/core_patternannotation is correctly set on the pod. This is a known issue in WRCP related to the coredump filter behavior for guaranteed pods. -
JCNR-13642: TX errors and control plane flaps occur when forwarded traffic exceeds 9.5 Gbps on GCP with gVNIC—When Cloud-Native Router is deployed as a transit router in GCP with gVNIC in GQI QPL mode and MPLSoUDP transport, TX errors occur on the PE-to-PE link when bidirectional forwarded traffic exceeds 9.5 Gbps. This is caused by a hypervisor limitation where the GVE hardware queues cannot dequeue packets fast enough, leading to NIC-level TX queue errors and possible BGP control plane flaps. As a workaround, configure a QoS scheduler profile with a bandwidth limit of 9 Gbps on the fabric interfaces to prevent control plane traffic from being dropped.