High memory usage

[thunderbird86]

Description

We faced high memory usage of cAdvisor on some nodes

There is no OOM kills, memory cleaned after some time, here is two pod on the same cluster

And spikes higher after 05/31, its just because we increased limits for pod, but it proceed consume all available memory

Details

Pods are placed on different node-groups separated by load type with affinity In other it's the same amd64 arch instances started from one image

cAdvisor versions which we tested v0.47.2 v0.49.1

kubectl version

Client Version: v1.29.2
Kustomize Version: v5.0.4-0.20230601165947-6ce0bf390ce3
Server Version: v1.27.13-eks-3af4770

cAdvisor params:

- --update_machine_info_interval=5m
- --housekeeping_interval=15s
- --max_housekeeping_interval=15s
- --event_storage_event_limit=default=0
- --event_storage_age_limit=default=0
- --containerd=/run/containerd/containerd.sock
- --disable_root_cgroup_stats=false
- --store_container_labels=false
- --whitelisted_container_labels=io.kubernetes.container.name,io.kubernetes.pod.name,io.kubernetes.pod.namespace
- --disable_metrics=percpu,tcp,udp,referenced_memory,advtcp,memory_numa,hugetlb
- --enable_load_reader=true
- --docker_only=true
- --profiling=true

go tool pprof -http=:8082 cadvisor-zd6t4-heap.out

[iwankgb]

@thunderbird86, can you try code from #3561?

[thunderbird86]

Thanks @iwankgb, I've rolled out it, and need couple of days to gather statistics

[thunderbird86]

Hi @iwankgb, I've tried few attempts, and it doesn't help Here is last week