Added: Node labels are now retrieved from the Kubernetes API and added to the
K8sNodeSample. They can now be selected in the Narrow Down Entities section of the infrastructure alerts page to filter entities when using the
K8sNodeSample, and can also be used on any NRQL statement when querying the
FROM K8sNodeSample SELECT average(cpuUsedCoreMilliseconds) WHERE `label.kubernetes.io/role` = 'master'
By default, Information retrieved from the Kubernetes API is cached for 5 minutes. The cache time can be changed with the
API_SERVER_CACHE_TTL environment variable.
Changed: The integration now uses the Infrastructure Agent v1.5.51. For more information, please refer to the infrastructure agent release notes between versions v1.5.31 and v1.5.51.
- Changed: The integration now uses the Infrastructure Agent v1.5.31. The biggest changes were major improvements to logging and to the StorageSampler. For more information, please refer to the infrastructure agent release notes between versions v1.3.18 and v1.5.31.
- Fixed: The unprivileged version of Kubernetes integration was running as root after a restart of the container due to kubernetes/kubernetes#78308.
- Fixed: Autodiscovery cache directory permissions got changed from
744in order to let the
nri-agentuser write inside. This change was necessary to release an unprivileged version of the Kubernetes integration.
- Added: The integration reports the name of the cluster as Infrastructure inventory.
- Added: The integration reports a new event type
K8sClusterSample. At this moment, these events contain only the cluster name as an attribute.
Added: Support for
kube-state-metrics (KSM) 1.5.
message attributes are added to
K8sPodSample. This will provide visibility into why a pod status is
For example, a pod that failed due to memory pressure on the node will report the following attributes:
Message: Pod The node was low on resource: [MemoryPressure]
It is possible to create an alert for any of these attributes.
memoryWorkingSetBytes attribute is added to
K8sContainerSample. This metric is used by the OOMkiller to decide when a container is using too much memory compared to its limit and should therefore be killed. It will enable more precise monitoring of the memory usage for containers.
Changed: Always request metrics from
kube-state-metrics in text format. In KSM's v1.5 this is the default regardless of the format requested.
namespaceName attribute was added to all the samples where the
namespace attribute is present. This is to align with standard naming for attributes. Other examples:
namespace attribute is deprecated. We recommend using