layout-3 layout-6 layout-1 layout-top
NVIDIA Docs Hubchevron_right NVIDIA Networkingchevron_right Networking Softwarechevron_right Management Softwarechevron_right

NVIDIA NetQ

layout-top layout-6 layout-1

A newer version of this product documentation is available. If you are redirected to the main page of the user guide, then this page might have been renamed or removed.

NVIDIA NetQ 5.2 Release Notes

Download 5.2 Release Notes xls    Download all 5.2 release notes as .xls

5.2.1 Release Notes

Open Issues in 5.2.1

Issue ID Description Affects Fixed
5073537 After uninstalling a NetQ cluster, the cluster VIP address is not released from the network interface. On a subsequent reinstall, the installer detects the VIP as still reachable and exits with the error cluster-vip address is reachable. Please provide non reachable IP address, even though the cluster is no longer running
This issue affects deployments that use a VLAN interface (for example, bond0.100) rather than the management IP for the cluster VIP. It does not affect first-time installs, but blocks every uninstall-reinstall cycle until the VIP is manually removed
To work around this issue, manually remove the VIP from the interface before reinstalling with sudo ip addr del / dev 		5.2.1
4986372 After installing NetQ for Ethernet and NVLink, the UI might fail to load and display the message No route matched with those values.. To work around this issue, restart the Kong-CP and Kong-DP deployments with kubectl rollout restart deployment -n netq-infra kong-cp-kongsleep 300, followed by kubectl rollout restart deployment -n netq-infra kong-dp-kong

If the issue persists, run cd /tmp/netq-infra/kube-config/build/data-infra, followed by kubectl apply -f deck-sync-job.yaml.		 5.2.1
4965251 When BlueField DPUs are in the default, embedded mode, the output of the netq check roce command might produce duplicate entries for the same device (one for the host and one for the DPU). This issue can also affect the calculations for the RoCE Mode Consistency and DSCP Classification tests. Additionally, the netq show roce-config host command might display DPU interfaces alongside host interfaces. This issue does not affect DPUs that are configured in separated mode. 5.2.1-5.3.0
4943571 After you register NVLink services, it might take up to an hour for the scheduler service to run and return data. 5.2.1
4867933 Threshold-crossing events created before version 5.1.0 may not display event values correctly after you upgrade NetQ. 5.1.0-5.2.1
4854663 When specifying a cluster VIP on an invalid or incorrect subnet, the installer displays an error indicating that master_ip should be different than cluster_ip without indicating which IP address is invalid. 5.1.0-5.2.1
4784336 The NetQ for NVLink deployment option is not supported in air-gapped environments. 5.0.0-5.2.1
4687477 When you run a validation against a group of devices with specific labels, NetQ ignores any pre-configured filters. 5.0.0-5.2.1
4682275 NVLink cluster installations do not validate that each node has a unique hostname. If two nodes share a common hostname, NetQ does not flag the issue after the installation completes. 5.0.0-5.2.1
4681581 The netq bootstrap reset purge-db command might take up to 60 minutes to complete on Base Command Manager scale deployments. 5.0.0-5.2.1
4399074 When connecting a switch to NMX-T or NMX-C through the service registration workflow, use either the IP address or the hostname. Using both creates duplicate registrations, and the operation does not fail as expected. 5.0.0-5.2.1
4389662 When a cluster installation fails a cluster VIP validation check, the installer generates an opta-support archive and prompts you to send it to NVIDIA support instead of prompting you to fix the initial error. 4.15.0-5.2.1
4122430 When the master node is unreachable, a worker node might report the output of netq show status as Not Installed instead of indicating that the cluster was degraded. 4.12.0-5.2.1
4100882, 4119697 When you attempt to export a file that is larger than 200MB, your browser might crash or otherwise prevent you from exporting the file. To work around this issue, use filters in the UI to decrease the size of the dataset that you intend to export. 4.12.0-4.15.1, 5.0.0-5.2.1

Fixed Issues in 5.2.1

Issue ID Description Affects
4977619 NetQ might fail to load in high-availability (HA) scale deployments when ECMP is enabled. To work around this issue, disable ECMP ingestion by updating the Kafka connector configuration:
  1. Create a backup of the current configuration:
    kubectl get kafkaconnector cassandra-sink-group-1 -n netq-infra -o yaml > /tmp/cassandra-sink-group-1.backup.yaml
  2. Capture the current Kafka connector values:
    OLD_KCQL=$(kubectl get kafkaconnector cassandra-sink-group-1 -n netq-infra -o jsonpath='{.spec.config.connect.cassandra.kcql}')
    OLD_TOPICS=$(kubectl get kafkaconnector cassandra-sink-group-1 -n netq-infra -o jsonpath='{.spec.config.topics}')
  3. Remove ECMP-related entries from KCQL and topics:
    NEW_KCQL=$(echo “$OLD_KCQL” | tr ‘;’ ‘\n’ | grep -v -E ‘INTO ecmp_info SELECT|INTO ecmp_info_aggregate SELECT|INTO ecmp_info_group_aggregate SELECT’ | paste -sd ‘;’ -)
    NEW_TOPICS=$(echo “$OLD_TOPICS” | tr ‘,’ ‘\n’ | grep -v -E ‘^netq_obj_ecmp_info$|^netq_obj_ecmp_info_aggregate$|^netq_obj_ecmp_info_group_aggregate$’ | paste -sd ‘,’ -)
  4. Verify the changes before applying them:
    echo “=== KCQL ECMP entries remaining (should only show ecmp_hash + arEcmpInfo) ==="
    echo “$NEW_KCQL” | tr ‘;’ ‘\n’ | grep -i ecmpecho “=== TOPICS ECMP entries remaining (should only show netq_obj_ecmp_hash) ==="
    echo “$NEW_TOPICS” | tr ‘,’ ‘\n’ | grep -i ecmp
    echo “Old KCQL: $(echo “$OLD_KCQL” | tr ‘;’ ‘\n’ | wc -l) / New: $(echo “$NEW_KCQL” | tr ‘;’ ‘\n’ | wc -l) (-3)"
    echo “Old TOPICS: $(echo “$OLD_TOPICS” | tr ‘,’ ‘\n’ | wc -l) / New: $(echo “$NEW_TOPICS” | tr ‘,’ ‘\n’ | wc -l) (-3)"
  5. Apply the updated configuration:
    kubectl patch kafkaconnector cassandra-sink-group-1 -n netq-infra –type=merge -p “$(jq -n –arg k “$NEW_KCQL” –arg t “$NEW_TOPICS” ‘{spec:{config:{“connect.cassandra.kcql”:$k,“topics”:$t}}}')"
To revert the changes, restore the backup with kubectl apply -f /tmp/cassandra-sink-group-1.backup.yaml
 5.1.0
4977342, 4889408 The NetQ UI might not display a complete list of interfaces. To work around this issue, restart the NetQ agent or upgrade to the latest NetQ version. 5.1.0
4964170 When you create a partition, the operation might fail with a BAD_PARAM error. To work around this issue, retry the operation until it is successful. 5.1.0
4897897 The NetQ NVLink API might fail to return a 400 error response when unsupported parameters are included in the API request. 5.1.0
4894672, 4896363 NVOS upgrades on switches might fail if the image filename contains spaces or special characters. 5.1.0
4890084 The NetQ CLI might not delete threshold-crossing alerts configured for ACL resources. To work around this issue, use the UI to delete the alerts. 5.1.0
4876932 The NMX controller service might intermittently fail to switch to an alternate out-of-band (OOB) port when the primary registration port becomes unavailable. As a result, partition management operations might not function correctly on the secondary OOB port if the registered management port goes down. 5.1.0
4839716, 4844441 You cannot perform lifecycle management operations immediately after backing up an HA scale cluster deployment. To work around this issue, wait several hours before initiating LCM operations. 5.1.0
4838526 NMX controller and telemetry services might display a DOWN status if the primary out-of-band (OOB) management port is unavailable. 5.1.0
4830357 When you try to filter images in the NetQ UI based on image type, NetQ might ignore the filter. 5.1.0
4794266 Power sensor (PSU) events might show inconsistent sensor names. Additionally, NetQ might not generate PSU events reliably. 5.1.0
4780773 The NetQ for NVLink Swagger UI might incorrectly display 500 status placeholder text in example responses. 5.0.0-5.1.0

    Copyright © 2026  NVIDIA Corporation