Labels for monitoring (Caascad)

Zone nomenclature for Caascad.

Dashboards

Note

There is no Grafana in Client cluster. No dashboard either.

Subzone name	Subzone	Label
Caascad	infra/staging infra/production cloud/caascad	grafana-dashboard-caascad
Cloud Client	cloud/client	grafana-dashboard-client

Note

The client has no access to K8S and will not be able to deploy configmaps directly. So the client has no knowledge of this label.

Datasources

Note

There is no Grafana in Client cluster. No datasouces either.

Subzone name	Subzone	Label
Caascad	infra/staging infra/production cloud/caascad	grafana-datasource-caascad="1"
Cloud Client	cloud/client	grafana-datasource-client="1"

podMonitors and serviceMonitors

Subzone name	Subzone	Label
Caascad	infra/staging infra/production cloud/caascad client/caascad	caascad.com/prometheus-monitor="caascad"
Consumption	infra/staging/consumption infra/production/consumption	caascad.com/prometheus-monitor="consumption"
Cloud Client	cloud/client	caascad.com/prometheus-monitor="client"
Cloud App	cloud/app	No podMonitors nor serviceMonitors in this zone
Dev	any subzone	caascad.com/prometheus-monitor="caascad"
Client	client	caascad.com/prometheus-monitor="caascad" + optional caascad.com/prometheus-monitor-scope="caascad" (see below)

Principle : if an application of the Cloud zone is exposed to the client, it will be monitored by the Prometheus in Cloud Client zone. Examples : git, CI/CD...

Client zone specific case

Client zones are scrapped Prometheus in the client zone. For this, label caascad.com/prometheus-monitor="caascad" is mandatory.

All metrics from Prometheus in the client zone will be stored in the Cloud-app zone.

If you want to duplicate the metrics of the client zone in the Cloud-client zone, you need to put an additionnal label caascad.com/prometheus-monitor-scope="caascad".

Note

Label caascad.com/prometheus-monitor-scope="app" is ignored. It used to be in use in an older architecture. It can safely be removed.

Dev environments specific case

You may set caascad.com/prometheus-monitor="caascad" like in the other environments. In fact, do what you want.

Additionnal labels to identify metrics, logs and alerts

See MVPMON-130 comments to see the long discussion how we chose those labels and their possible values.

Some additionnal labels may be needed to identify a metric :

• where does it comes from (which prometheus scraped it) ?
• where is it stored ?

Those labels may be added to better identify a metric :

• cc_prom : prometheus where the metric is used (visualization with grafana and alert generation)
• cc_prom_source : prometheus where the metric comes from
• obs_client (default) and/or cc_client : identity of the client
• cc_consumption_client : identity of the billed client

Label cc_prom : possible values

Prometheus	Label value
infra/caascad	infra-caascad
infra/consumption	infra-consumption
cloud/caascad	cloud-caascad
cloud/client	cloud-client
cloud/app	cloud-app
client	N/A

Use-case : for alerting

Label cc_prom_source : possible values

Same as cc_prom above.

For Prometheus in client zone, value is <cluster ID>.

Use-case : for logging, monitoring and alerting

Label obs_client and/or cc_client : possible values

Value is <client ID> except in infra zone.

In infra zone, value is infra-stg or infra-prd.

Use-case : for alerting

Label cc_consumption_client : possible values

Used only in infra zone.

Value is <client ID>.

Need and override for labels cc_* and/or obs_*

label	tool	need	override is possible ?
cc_prom	metrics	optional	yes
cc_prom	logs	optional	yes
cc_prom	alerts	mandatory	yes
cc_prom_source	metrics	mandatory (see note below)	no (see note below)
cc_prom_source	logs	mandatory	no
cc_prom_source	alerts	mandatory	no
cc_client	metrics	optional	no
cc_client	logs	optional	no
cc_client	alerts	mandatory	no
obs_client	metrics	optional	no
obs_client	logs	optional	no
obs_client	alerts	mandatory	no
cc_consumption_client	metrics	mandatory	no
cc_consumption_client	logs	no	no
cc_consumption_client	alerts	mandatory	no

Note

In Prometheus cloud-client and cloud-app, the cc_prom_source is set (or overwritten) at the federation step because we do not expect the client to set this label himself.

PrometheusRules

Note

There is no Alertmanager in Client cluster. No prometheusrules either.

Subzone name	Subzone	Label
Caascad	infra/staging infra/production cloud/caascad	caascad.com/prometheus-rules: "caascad"
Consumption	infra/staging/consumption infra/production/consumption	caascad.com/prometheus-rules: "consumption"
Cloud Client	cloud/client	caascad.com/prometheus-rules: "client"
Cloud App	cloud/app	caascad.com/prometheus-rules: "app"