Kubernetes deployment metrics calculate is wrong
Closed this issue · 2 comments
guitcastro commented
I have two deployments foo-api
and foo-api-stg
, when aggregating data from foo-api
metrics from foo-api-stg
are summed together. This is because of the regex used in the dashboard:
sum (container_memory_working_set_bytes{image!="",name=~"^k8s_.*",pod_name=~"^$Deployment.*$",container_name!="POD"}) by (pod_name)
Anyway thoughts on how truly separate the pod metrics?
Thakurvaibhav commented
Hi @guitcastro
Thanks for reaching out.
You can do multiple things here:
- Completely remove the
Deployment
from scope and group by pods only. - Add
Namespace
as variable in grafana just like theDeployment
and then filter using namespace as well.
From the information you have provided it is not clear whether the deployments are in same namespace or not but it is highly recommended that deployments of same app for diff env should be in different namespaces if not different clusters. However, if you have 2 deployments of similar name in the same namespace, I would simply update the query to group by pods and ignore the deployment name altogether.
Does that sound viable ?
Thakurvaibhav commented
@guitcastro I believe you fixed it. Closing it. Let me know if you want me to re-open.