honor_timestamps parameter usage

[Loic Fouray]

Hi,

I have the following use case :

- I use a cadvisor job that scraps the system metrics on the Kubernetes nodes :

- job_name: node/cadvisor

honor_labels: true

honor_timestamps: true

scrape_interval: 10s

scrape_timeout: 10s

metrics_path: /metrics/cadvisor

scheme: https

kubernetes_sd_configs:

- role: node

namespaces:

names:

- <namespace>

bearer_token_file: /var/run/secrets/kubernetes.io/serviceaccount/token

tls_config:

ca_file: /var/run/secrets/kubernetes.io/serviceaccount/ca.crt

insecure_skip_verify: true

- I use the container_memory_working_set_bytes metrics to retrieve the RAM used by my containers.

- When i redeploy my pod, i have 2 sample by promql  'instant query" (without the time parameter). Indeed, the sample related to the previous pod/deployment is still returned during few minutes,

- If i set honor_timestamps to false instead true, i have only the last sample active, it works.

Could you more information on this parameter?

Thanks

Loïc

[Loïc]

Hi,

For my use case, the usage of honor_timestamps parameter with false value is ok for my need. But, i search to know what the other impacts can be. Have you an idea?

Loïc