-
Story
-
Resolution: Done
-
Undefined
-
None
-
None
-
None
-
5
-
False
-
None
-
False
-
-
-
NetObserv - Sprint 238, NetObserv - Sprint 239, NetObserv - Sprint 240, NetObserv - Sprint 245, NetObserv - Sprint 246, NetObserv - Sprint 247, NetObserv - Sprint 248
Create new metrics to report top pod talkers as:
- direction (egress / ingess)
- kind (pod / service)
- name
- namespace
- node name
- owner name
- bytes
To create a new metric, a definition file must be added in https://github.com/netobserv/network-observability-operator/tree/main/controllers/flowlogspipeline/metrics_definitions , similar to the ones already there.
For this metric, I would imagine something along those lines could work (not tested) :
encode:
type: prom
prom:
metrics:
- name: top_pods_egress_bytes_total
type: counter
valuekey: Bytes
filter:
key: FlowDirection
value: "1"
labels:
- SrcK8S_Name
- SrcK8S_Namespace
- SrcK8S_Kind
- SrcK8S_Node
And another one by destinations filtered on ingress
I don't remember how to do the "Top" thing but I remember it's possible 
- blocks
-
-
- Closed
-
- is blocked by
-
-
- Closed
-
-
NETOBSERV-1131 Metrics do not ignore duplicates
-
- Closed
-
-
-
- Closed
-
- links to
- mentioned on
[NETOBSERV-1017] Metrics: Top pod or service talkers
GitLab CEE Bot
added a comment - CPaaS Service Account mentioned this issue in a merge request of netobserv-midstream / Network Observability Operator Midstream on branch network-observability-1.5.0-rhel-9_upstream_f8125d8ea89d35ae383abef4ee6d396c:
Updated 2 upstream sources
GitLab CEE Bot
added a comment - CPaaS Service Account mentioned this issue in a merge request of netobserv-midstream / Network Observability Operator Midstream on branch network-observability-1.5.0-rhel-9_ upstream _f8125d8ea89d35ae383abef4ee6d396c : Updated 2 upstream sources GitLab CEE Bot
added a comment - CPaaS Service Account mentioned this issue in a merge request of netobserv-midstream / Network Observability Operator Midstream on branch network-observability-1.5.0-rhel-9_upstream_0c7c21eea8e584cde154174f75630e7a:
Updated US source to: da4bdd3
NETOBSERV-1017: pods/services bandwidth (#523)
GitLab CEE Bot
added a comment - CPaaS Service Account mentioned this issue in a merge request of netobserv-midstream / Network Observability Operator Midstream on branch network-observability-1.5.0-rhel-9_ upstream _0c7c21eea8e584cde154174f75630e7a : Updated US source to: da4bdd3 NETOBSERV-1017 : pods/services bandwidth (#523) This is part of the dev preview epic for custom metrics API, so marking no-doc no-qe ; moreover, this is only providing sample YAMLs in the repository: nothing impacting the operator directly.
On this story: playing with the new metrics API I'm definitely able to create that metric. But thinking twice, I don't think it brings any value on top of the existing node-exporter metrics, which provide the exact same thing if I'm correct. So we could just do that as an example, without actually deploying it?
Following jtakvori feedback I removed the topk stage since the impact seems to be lower than expected.
We need to run perfscale to ensure the observed impact is acceptable before merging that PR
GitLab CEE Bot
added a comment - CPaaS Service Account mentioned this issue in a merge request of netobserv-midstream / Network Observability Operator Midstream on branch network-observability-1.4.0-rhel-9_upstream_ce48126351e85b21efe55798720286eb:
Updated 2 upstream sources
GitLab CEE Bot
added a comment - CPaaS Service Account mentioned this issue in a merge request of netobserv-midstream / Network Observability Operator Midstream on branch network-observability-1.4.0-rhel-9_ upstream _ce48126351e85b21efe55798720286eb : Updated 2 upstream sources
CPaaS Service Account mentioned this issue in a merge request of netobserv-midstream / Network Observability Operator Midstream on branch network-observability-1.5.0-rhel-9_upstream_1b0121875cedab01b994cb3bafde63a9: