Hello team . There is an inconsistency in reporting of memory usage for pod on openobserve and kubectl top command Value reported by kubectl top command: quilradmin@Redacted ~]$ kubectl top pod kafka-server-0 -n quilr NAME CPU(cores) MEMORY(bytes) kafka-server-0 34m 2232Mi Value reported by openosberve:

{
__hash__: 12312006193863854000,
__name__: k8s_pod_memory_usage,
_timestamp: 1783470632364408,
flag: DATA_POINT_FLAGS_DO_NOT_USE,
host_name: o2c-openobserve-collector-agent-collector-jh9vs,
instrumentation_library_name: github.com/open-telemetry/opentelemetry-collector-contrib/receiver/kubeletstatsreceiver,
instrumentation_library_version: 0.138.0,
k8s_cluster: quilr-production-useast,
k8s_container_name: kafka,
k8s_namespace_name: quilr,
k8s_node_name: ip-10-81-17-29.ec2.internal,
k8s_node_uid: ec3fe0aa-d5d3-480d-91fa-3ef2605c50f7,
k8s_pod_name: kafka-server-0,
k8s_pod_start_time: 2026-05-27T09:30:24Z,
k8s_pod_uid: cdd85329-eb10-45f8-8234-490e210adecd,
os_type: linux,
start_time: 1779874225000000000,
value: 4294643712
}

These false positives are pretty common for services in the cluster . we are using openobserve v0.80.3

Hey @Varun Manjunath let us quickly check and revert back on this!

this is not about OpenObserve. we are only accept data, store data, query data.

@Hengfei got the issue basically need to use k8s_pod_memory_working_set gives right result. Thank you for reference