diff --git a/.gitignore b/.gitignore
new file mode 100644
index 0000000000000000000000000000000000000000..485dee64bcfb48793379b200a1afd14e85a8aaf4
--- /dev/null
+++ b/.gitignore
@@ -0,0 +1 @@
+.idea
diff --git a/Dockerfile b/Dockerfile
index c62e9cb8260b3159f543ee714fc8a3e9588edf2d..85c567fd969a0828cc2daafddf1cffb2cbadef77 100644
--- a/Dockerfile
+++ b/Dockerfile
@@ -5,6 +5,8 @@ USER root
RUN apk --no-cache add curl
RUN apk --no-cache add jq
+RUN wget https://github.com/mikefarah/yq/releases/download/v4.44.1/yq_linux_amd64.tar.gz -O - |\
+ tar xz && mv yq_linux_amd64 /usr/bin/yq && chmod +x /usr/bin/yq
USER grafana
@@ -21,15 +23,16 @@ RUN grafana cli plugins install grafana-oncall-app
RUN wget https://algenty.github.io/flowcharting-repository/archives/agenty-flowcharting-panel-1.0.0b-SNAPSHOT.zip -O /tmp/agenty-flowcharting-panel.zip
RUN cd /var/lib/grafana/plugins/ && unzip /tmp/agenty-flowcharting-panel.zip && mv grafana-flowcharting agenty-flowcharting-panel
-COPY grafana.ini /etc/grafana/
-COPY alerting.json /opt/grafana-import/
-COPY rules.json /opt/grafana-import/
-COPY import-rules.sh /opt/grafana-import/
+COPY grafana.ini /etc/grafana/grafana.ini
+COPY imports/populate-tokens.sh /opt/grafana-import/populate-tokens.sh
# Add default configuration through provisioning (see https://grafana.com/docs/grafana/latest/administration/provisioning)
+# https://grafana.com/docs/grafana/latest/alerting/set-up/provision-alerting-resources/file-provisioning/
+COPY alerting /etc/grafana/provisioning/alerting/
COPY datasources /etc/grafana/provisioning/datasources/
-COPY dashboards /var/lib/grafana/dashboards/
-COPY panels /var/lib/grafana/panels/
+COPY station-dashboards.yaml /etc/grafana/provisioning/dashboards/
+COPY dashboards /var/lib/grafana/dashboards/station/station/
+COPY panels /var/lib/grafana/panels/panels/
COPY run-wrapper.sh /run-wrapper.sh
ENTRYPOINT ["/run-wrapper.sh"]
diff --git a/README.md b/README.md
index b66b780f146abf242577e70dffd4034cd9f8f640..d6872777ecbb9fea28ad888801fda6f30a077acb 100644
--- a/README.md
+++ b/README.md
@@ -2,12 +2,19 @@
These dashboards show the state of a single station. They are tailored to be installed:
-* Locally on a station, through https://git.astron.nl/lofar2.0/tango/-/tree/master/docker-compose/grafana
+* Locally on a station, through https://git.astron.nl/lofar2.0/tango/-/tree/master/docker/grafana
* Centrally to monitor a group of stations, through https://git.astron.nl/lofar2.0/operations-central-management/-/tree/main/grafana-central/dashboards
+## Environment variables
+
+This container uses the following environment variables upon starting for configuration
+
+1. `SLACK_TOKEN`: Authentication token for slack alarms
+
## Datasources
-The Grafana installation in this repo exposes and uses the following data sources, as configured in the `dashboards/` directory:
+The Grafana installation in this repo exposes and uses the following data sources,
+as configured in the `dashboards/` directory:
* Prometheus, at `http://prometheus:9090`, serving metrics,
* Loki, at `http://loki:3100`, serving logs,
@@ -25,6 +32,6 @@ To cover both these use cases, the designer must consider:
To deploy changes, they must be:
-1. Committed to this repository,
+1. Commit to this repository,
2. The submodules in the tango and operations-central-management repositories must link to the new commit,
3. Those repositories need to be redeployed on the stations and centrally, respectively.
diff --git a/alerting.json b/alerting.json
deleted file mode 100644
index bc5c76e7f8870efa52e60e21bf621ae0f1cd8418..0000000000000000000000000000000000000000
--- a/alerting.json
+++ /dev/null
@@ -1,26 +0,0 @@
-{
- "template_files": {},
- "alertmanager_config": {
- "route": {
- "receiver": "Alerta",
- "repeat_interval": "10m"
- },
- "templates": null,
- "receivers": [
- {
- "name": "Alerta",
- "grafana_managed_receiver_configs": [
- {
- "name": "Alerta",
- "type": "webhook",
- "disableResolveMessage": false,
- "settings": {
- "url": "http://alerta-server:8080/api/webhooks/prometheus?api-key=demo-key"
- },
- "secureFields": {}
- }
- ]
- }
- ]
- }
-}
diff --git a/alerting/alert-rules.yaml b/alerting/alert-rules.yaml
new file mode 100644
index 0000000000000000000000000000000000000000..e2039f97feea0202987d46cfcfbed922400c7046
--- /dev/null
+++ b/alerting/alert-rules.yaml
@@ -0,0 +1,1667 @@
+apiVersion: 1
+groups:
+ - orgId: 1
+ name: Evaluate every minte
+ folder: station
+ interval: 1m
+ rules:
+ - uid: edloik9y36m0we
+ title: FPGA Temperatures hba0
+ condition: C
+ data:
+ - refId: B
+ relativeTimeRange:
+ from: 1800
+ to: 0
+ datasourceUid: prometheus
+ model:
+ datasource:
+ type: prometheus
+ uid: prometheus
+ editorMode: code
+ expr: ds_fpga_temp{host="localhost",family="SDPFirmware", member="HBA0"} != 0
+ instant: false
+ interval: ""
+ intervalMs: 15000
+ legendFormat: '{{x}}'
+ maxDataPoints: 43200
+ range: true
+ refId: B
+ - refId: A
+ relativeTimeRange:
+ from: 1800
+ to: 0
+ datasourceUid: __expr__
+ model:
+ conditions:
+ - evaluator:
+ params: []
+ type: gt
+ operator:
+ type: and
+ query:
+ params:
+ - A
+ reducer:
+ params: []
+ type: last
+ type: query
+ datasource:
+ type: __expr__
+ uid: __expr__
+ expression: B
+ intervalMs: 1000
+ maxDataPoints: 43200
+ reducer: max
+ refId: A
+ type: reduce
+ - refId: C
+ relativeTimeRange:
+ from: 1800
+ to: 0
+ datasourceUid: __expr__
+ model:
+ conditions:
+ - evaluator:
+ params:
+ - 85
+ type: gt
+ operator:
+ type: and
+ query:
+ params:
+ - C
+ reducer:
+ params: []
+ type: last
+ type: query
+ datasource:
+ type: __expr__
+ uid: __expr__
+ expression: A
+ intervalMs: 1000
+ maxDataPoints: 43200
+ refId: C
+ type: threshold
+ dashboardUid: c54bff42-4ff7-4e1d-8bf0-1f1c9cc0ea35
+ panelId: 6
+ noDataState: NoData
+ execErrState: Error
+ for: 5m
+ annotations:
+ __dashboardUid__: c54bff42-4ff7-4e1d-8bf0-1f1c9cc0ea35
+ __panelId__: "6"
+ description: The FPGA temperature in hba0 is to high
+ summary: FPGA temperature in hba0 have been above alarm conditions for the last 5 minutes.
+ labels: {}
+ isPaused: false
+ - uid: fdloko30iiv40a
+ title: FPGA Temperatures hba1
+ condition: C
+ data:
+ - refId: B
+ relativeTimeRange:
+ from: 1800
+ to: 0
+ datasourceUid: prometheus
+ model:
+ datasource:
+ type: prometheus
+ uid: prometheus
+ editorMode: code
+ expr: ds_fpga_temp{host="localhost",family="SDPFirmware", member="HBA1"} != 0
+ instant: false
+ interval: ""
+ intervalMs: 15000
+ legendFormat: '{{x}}'
+ maxDataPoints: 43200
+ range: true
+ refId: B
+ - refId: A
+ relativeTimeRange:
+ from: 1800
+ to: 0
+ datasourceUid: __expr__
+ model:
+ conditions:
+ - evaluator:
+ params: []
+ type: gt
+ operator:
+ type: and
+ query:
+ params:
+ - A
+ reducer:
+ params: []
+ type: last
+ type: query
+ datasource:
+ type: __expr__
+ uid: __expr__
+ expression: B
+ intervalMs: 1000
+ maxDataPoints: 43200
+ reducer: max
+ refId: A
+ type: reduce
+ - refId: C
+ relativeTimeRange:
+ from: 1800
+ to: 0
+ datasourceUid: __expr__
+ model:
+ conditions:
+ - evaluator:
+ params:
+ - 85
+ type: gt
+ operator:
+ type: and
+ query:
+ params:
+ - C
+ reducer:
+ params: []
+ type: last
+ type: query
+ datasource:
+ type: __expr__
+ uid: __expr__
+ expression: A
+ intervalMs: 1000
+ maxDataPoints: 43200
+ refId: C
+ type: threshold
+ dashboardUid: c54bff42-4ff7-4e1d-8bf0-1f1c9cc0ea35
+ panelId: 6
+ noDataState: NoData
+ execErrState: Error
+ for: 5m
+ annotations:
+ __dashboardUid__: c54bff42-4ff7-4e1d-8bf0-1f1c9cc0ea35
+ __panelId__: "6"
+ summary: FPGA temperature in hba1 have been above alarm conditions for the last 5 minutes.
+ labels: {}
+ isPaused: false
+ - uid: adlokt8acedj4c
+ title: FPGA Temperatures lba
+ condition: C
+ data:
+ - refId: B
+ relativeTimeRange:
+ from: 1800
+ to: 0
+ datasourceUid: prometheus
+ model:
+ datasource:
+ type: prometheus
+ uid: prometheus
+ editorMode: code
+ expr: ds_fpga_temp{host="localhost",family="SDPFirmware", member="LBA"} != 0
+ instant: false
+ interval: ""
+ intervalMs: 15000
+ legendFormat: '{{x}}'
+ maxDataPoints: 43200
+ range: true
+ refId: B
+ - refId: A
+ relativeTimeRange:
+ from: 1800
+ to: 0
+ datasourceUid: __expr__
+ model:
+ conditions:
+ - evaluator:
+ params: []
+ type: gt
+ operator:
+ type: and
+ query:
+ params:
+ - A
+ reducer:
+ params: []
+ type: last
+ type: query
+ datasource:
+ type: __expr__
+ uid: __expr__
+ expression: B
+ intervalMs: 1000
+ maxDataPoints: 43200
+ reducer: max
+ refId: A
+ type: reduce
+ - refId: C
+ relativeTimeRange:
+ from: 1800
+ to: 0
+ datasourceUid: __expr__
+ model:
+ conditions:
+ - evaluator:
+ params:
+ - 85
+ type: gt
+ operator:
+ type: and
+ query:
+ params:
+ - C
+ reducer:
+ params: []
+ type: last
+ type: query
+ datasource:
+ type: __expr__
+ uid: __expr__
+ expression: A
+ intervalMs: 1000
+ maxDataPoints: 43200
+ refId: C
+ type: threshold
+ dashboardUid: c54bff42-4ff7-4e1d-8bf0-1f1c9cc0ea35
+ panelId: 6
+ noDataState: NoData
+ execErrState: Error
+ for: 5m
+ annotations:
+ __dashboardUid__: c54bff42-4ff7-4e1d-8bf0-1f1c9cc0ea35
+ __panelId__: "6"
+ summary: FPGA temperature in lba have been above alarm conditions for the last 5 minutes.
+ labels: {}
+ isPaused: false
+ - uid: fdlon8p6bvri8e
+ title: Uniboard2 FPGA POL Core Temperatures H0
+ condition: H
+ data:
+ - refId: A
+ relativeTimeRange:
+ from: 1800
+ to: 0
+ datasourceUid: prometheus
+ model:
+ datasource:
+ type: prometheus
+ uid: prometheus
+ editorMode: code
+ exemplar: true
+ expr: 'ds_unb2_fpga_pol_core_temp{host="localhost",family="UNB2",member="H0"} '
+ instant: false
+ interval: ""
+ intervalMs: 15000
+ legendFormat: Core board {{x}} node {{y}}
+ maxDataPoints: 43200
+ range: true
+ refId: A
+ - refId: G
+ relativeTimeRange:
+ from: 1800
+ to: 0
+ datasourceUid: __expr__
+ model:
+ conditions:
+ - evaluator:
+ params: []
+ type: gt
+ operator:
+ type: and
+ query:
+ params:
+ - G
+ reducer:
+ params: []
+ type: last
+ type: query
+ datasource:
+ type: __expr__
+ uid: __expr__
+ expression: A
+ intervalMs: 1000
+ maxDataPoints: 43200
+ reducer: max
+ refId: G
+ settings:
+ mode: dropNN
+ type: reduce
+ - refId: H
+ relativeTimeRange:
+ from: 1800
+ to: 0
+ datasourceUid: __expr__
+ model:
+ conditions:
+ - evaluator:
+ params:
+ - 85
+ type: gt
+ operator:
+ type: and
+ query:
+ params:
+ - H
+ reducer:
+ params: []
+ type: last
+ type: query
+ datasource:
+ type: __expr__
+ uid: __expr__
+ expression: G
+ intervalMs: 1000
+ maxDataPoints: 43200
+ refId: H
+ type: threshold
+ dashboardUid: c54bff42-4ff7-4e1d-8bf0-1f1c9cc0ea35
+ panelId: 4
+ noDataState: NoData
+ execErrState: Error
+ for: 5m
+ annotations:
+ __dashboardUid__: c54bff42-4ff7-4e1d-8bf0-1f1c9cc0ea35
+ __panelId__: "4"
+ description: FPGA POL Core temperature for H0 has been above alarm conditions for 5 minutes.
+ summary: FPGA POL Core temperature for H0 has been above alarm conditions for 5 minutes.
+ labels: {}
+ isPaused: false
+ - uid: edlonaz69vy80d
+ title: Uniboard2 FPGA POL Core Temperatures L0
+ condition: H
+ data:
+ - refId: A
+ relativeTimeRange:
+ from: 1800
+ to: 0
+ datasourceUid: prometheus
+ model:
+ datasource:
+ type: prometheus
+ uid: prometheus
+ editorMode: code
+ exemplar: true
+ expr: 'ds_unb2_fpga_pol_core_temp{host="localhost",family="UNB2",member="L0"} '
+ instant: false
+ interval: ""
+ intervalMs: 15000
+ legendFormat: Core board {{x}} node {{y}}
+ maxDataPoints: 43200
+ range: true
+ refId: A
+ - refId: G
+ relativeTimeRange:
+ from: 1800
+ to: 0
+ datasourceUid: __expr__
+ model:
+ conditions:
+ - evaluator:
+ params: []
+ type: gt
+ operator:
+ type: and
+ query:
+ params:
+ - G
+ reducer:
+ params: []
+ type: last
+ type: query
+ datasource:
+ type: __expr__
+ uid: __expr__
+ expression: A
+ intervalMs: 1000
+ maxDataPoints: 43200
+ reducer: max
+ refId: G
+ settings:
+ mode: dropNN
+ type: reduce
+ - refId: H
+ relativeTimeRange:
+ from: 1800
+ to: 0
+ datasourceUid: __expr__
+ model:
+ conditions:
+ - evaluator:
+ params:
+ - 85
+ type: gt
+ operator:
+ type: and
+ query:
+ params:
+ - H
+ reducer:
+ params: []
+ type: last
+ type: query
+ datasource:
+ type: __expr__
+ uid: __expr__
+ expression: G
+ intervalMs: 1000
+ maxDataPoints: 43200
+ refId: H
+ type: threshold
+ dashboardUid: c54bff42-4ff7-4e1d-8bf0-1f1c9cc0ea35
+ panelId: 4
+ noDataState: NoData
+ execErrState: Error
+ for: 5m
+ annotations:
+ __dashboardUid__: c54bff42-4ff7-4e1d-8bf0-1f1c9cc0ea35
+ __panelId__: "4"
+ description: FPGA POL Core temperature for L0 has been above alarm conditions for 5 minutes.
+ summary: FPGA POL Core temperature for L0 has been above alarm conditions for 5 minutes.
+ labels: {}
+ isPaused: false
+ - uid: edlondwb8j474a
+ title: Uniboard2 FPGA POL Core Temperatures L1
+ condition: H
+ data:
+ - refId: A
+ relativeTimeRange:
+ from: 1800
+ to: 0
+ datasourceUid: prometheus
+ model:
+ datasource:
+ type: prometheus
+ uid: prometheus
+ editorMode: code
+ exemplar: true
+ expr: 'ds_unb2_fpga_pol_core_temp{host="localhost",family="UNB2",member="L1"} '
+ instant: false
+ interval: ""
+ intervalMs: 15000
+ legendFormat: Core board {{x}} node {{y}}
+ maxDataPoints: 43200
+ range: true
+ refId: A
+ - refId: G
+ relativeTimeRange:
+ from: 1800
+ to: 0
+ datasourceUid: __expr__
+ model:
+ conditions:
+ - evaluator:
+ params: []
+ type: gt
+ operator:
+ type: and
+ query:
+ params:
+ - G
+ reducer:
+ params: []
+ type: last
+ type: query
+ datasource:
+ type: __expr__
+ uid: __expr__
+ expression: A
+ intervalMs: 1000
+ maxDataPoints: 43200
+ reducer: max
+ refId: G
+ settings:
+ mode: dropNN
+ type: reduce
+ - refId: H
+ relativeTimeRange:
+ from: 1800
+ to: 0
+ datasourceUid: __expr__
+ model:
+ conditions:
+ - evaluator:
+ params:
+ - 85
+ type: gt
+ operator:
+ type: and
+ query:
+ params:
+ - H
+ reducer:
+ params: []
+ type: last
+ type: query
+ datasource:
+ type: __expr__
+ uid: __expr__
+ expression: G
+ intervalMs: 1000
+ maxDataPoints: 43200
+ refId: H
+ type: threshold
+ dashboardUid: c54bff42-4ff7-4e1d-8bf0-1f1c9cc0ea35
+ panelId: 4
+ noDataState: NoData
+ execErrState: Error
+ for: 5m
+ annotations:
+ __dashboardUid__: c54bff42-4ff7-4e1d-8bf0-1f1c9cc0ea35
+ __panelId__: "4"
+ description: FPGA POL Core temperature for L1 has been above alarm conditions for 5 minutes.
+ summary: FPGA POL Core temperature for L1 has been above alarm conditions for 5 minutes.
+ labels: {}
+ isPaused: false
+ - uid: adlonqgt6crgga
+ title: Uniboard2 FPGA POL ERAM Temperatures H0
+ condition: H
+ data:
+ - refId: B
+ relativeTimeRange:
+ from: 1800
+ to: 0
+ datasourceUid: prometheus
+ model:
+ datasource:
+ type: prometheus
+ uid: prometheus
+ editorMode: code
+ exemplar: true
+ expr: ds_unb2_fpga_pol_eram_temp{host="localhost",family="UNB2",member="H0"}
+ interval: ""
+ intervalMs: 15000
+ legendFormat: ERAM board {{x}} node {{y}}
+ maxDataPoints: 43200
+ range: true
+ refId: B
+ - refId: G
+ relativeTimeRange:
+ from: 1800
+ to: 0
+ datasourceUid: __expr__
+ model:
+ conditions:
+ - evaluator:
+ params: []
+ type: gt
+ operator:
+ type: and
+ query:
+ params:
+ - G
+ reducer:
+ params: []
+ type: last
+ type: query
+ datasource:
+ type: __expr__
+ uid: __expr__
+ expression: B
+ intervalMs: 1000
+ maxDataPoints: 43200
+ reducer: max
+ refId: G
+ settings:
+ mode: dropNN
+ type: reduce
+ - refId: H
+ relativeTimeRange:
+ from: 1800
+ to: 0
+ datasourceUid: __expr__
+ model:
+ conditions:
+ - evaluator:
+ params:
+ - 85
+ type: gt
+ operator:
+ type: and
+ query:
+ params:
+ - H
+ reducer:
+ params: []
+ type: last
+ type: query
+ datasource:
+ type: __expr__
+ uid: __expr__
+ expression: G
+ intervalMs: 1000
+ maxDataPoints: 43200
+ refId: H
+ type: threshold
+ dashboardUid: c54bff42-4ff7-4e1d-8bf0-1f1c9cc0ea35
+ panelId: 4
+ noDataState: NoData
+ execErrState: Error
+ for: 5m
+ annotations:
+ __dashboardUid__: c54bff42-4ff7-4e1d-8bf0-1f1c9cc0ea35
+ __panelId__: "4"
+ description: FPGA POL ERAM temperature for H0 has been above alarm conditions for 5 minutes.
+ summary: FPGA POL ERAM temperature for H0 has been above alarm conditions for 5 minutes.
+ labels: {}
+ isPaused: false
+ - uid: bdlonzk67asxsb
+ title: Uniboard2 FPGA POL ERAM Temperatures L0
+ condition: H
+ data:
+ - refId: B
+ relativeTimeRange:
+ from: 1800
+ to: 0
+ datasourceUid: prometheus
+ model:
+ datasource:
+ type: prometheus
+ uid: prometheus
+ editorMode: code
+ exemplar: true
+ expr: ds_unb2_fpga_pol_eram_temp{host="localhost",family="UNB2",member="L0"}
+ instant: false
+ interval: ""
+ intervalMs: 15000
+ legendFormat: ERAM board {{x}} node {{y}}
+ maxDataPoints: 43200
+ range: true
+ refId: B
+ - refId: G
+ relativeTimeRange:
+ from: 1800
+ to: 0
+ datasourceUid: __expr__
+ model:
+ conditions:
+ - evaluator:
+ params: []
+ type: gt
+ operator:
+ type: and
+ query:
+ params:
+ - G
+ reducer:
+ params: []
+ type: last
+ type: query
+ datasource:
+ type: __expr__
+ uid: __expr__
+ expression: B
+ intervalMs: 1000
+ maxDataPoints: 43200
+ reducer: max
+ refId: G
+ settings:
+ mode: dropNN
+ type: reduce
+ - refId: H
+ relativeTimeRange:
+ from: 1800
+ to: 0
+ datasourceUid: __expr__
+ model:
+ conditions:
+ - evaluator:
+ params:
+ - 85
+ type: gt
+ operator:
+ type: and
+ query:
+ params:
+ - H
+ reducer:
+ params: []
+ type: last
+ type: query
+ datasource:
+ type: __expr__
+ uid: __expr__
+ expression: G
+ intervalMs: 1000
+ maxDataPoints: 43200
+ refId: H
+ type: threshold
+ dashboardUid: c54bff42-4ff7-4e1d-8bf0-1f1c9cc0ea35
+ panelId: 4
+ noDataState: NoData
+ execErrState: Error
+ for: 5m
+ annotations:
+ __dashboardUid__: c54bff42-4ff7-4e1d-8bf0-1f1c9cc0ea35
+ __panelId__: "4"
+ description: FPGA POL ERAM temperature for L0 has been above alarm conditions for 5 minutes.
+ summary: FPGA POL ERAM temperature for L0 has been above alarm conditions for 5 minutes.
+ labels: {}
+ isPaused: false
+ - uid: fdloo6q2sdxq8d
+ title: Uniboard2 FPGA POL ERAM Temperatures L1
+ condition: H
+ data:
+ - refId: B
+ relativeTimeRange:
+ from: 1800
+ to: 0
+ datasourceUid: prometheus
+ model:
+ datasource:
+ type: prometheus
+ uid: prometheus
+ editorMode: code
+ exemplar: true
+ expr: ds_unb2_fpga_pol_eram_temp{host="localhost",family="UNB2",member="L1"}
+ instant: false
+ interval: ""
+ intervalMs: 15000
+ legendFormat: ERAM board {{x}} node {{y}}
+ maxDataPoints: 43200
+ range: true
+ refId: B
+ - refId: G
+ relativeTimeRange:
+ from: 1800
+ to: 0
+ datasourceUid: __expr__
+ model:
+ conditions:
+ - evaluator:
+ params: []
+ type: gt
+ operator:
+ type: and
+ query:
+ params:
+ - G
+ reducer:
+ params: []
+ type: last
+ type: query
+ datasource:
+ type: __expr__
+ uid: __expr__
+ expression: B
+ intervalMs: 1000
+ maxDataPoints: 43200
+ reducer: max
+ refId: G
+ settings:
+ mode: dropNN
+ type: reduce
+ - refId: H
+ relativeTimeRange:
+ from: 1800
+ to: 0
+ datasourceUid: __expr__
+ model:
+ conditions:
+ - evaluator:
+ params:
+ - 85
+ type: gt
+ operator:
+ type: and
+ query:
+ params:
+ - H
+ reducer:
+ params: []
+ type: last
+ type: query
+ datasource:
+ type: __expr__
+ uid: __expr__
+ expression: G
+ intervalMs: 1000
+ maxDataPoints: 43200
+ refId: H
+ type: threshold
+ dashboardUid: c54bff42-4ff7-4e1d-8bf0-1f1c9cc0ea35
+ panelId: 4
+ noDataState: NoData
+ execErrState: Error
+ for: 5m
+ annotations:
+ __dashboardUid__: c54bff42-4ff7-4e1d-8bf0-1f1c9cc0ea35
+ __panelId__: "4"
+ description: FPGA POL ERAM temperature for L1 has been above alarm conditions for 5 minutes.
+ summary: FPGA POL ERAM temperature for L1 has been above alarm conditions for 5 minutes.
+ labels: {}
+ isPaused: false
+ - uid: edlooih1cfpq8d
+ title: Uniboard2 FPGA POL rxgxb Temperatures H0
+ condition: H
+ data:
+ - refId: C
+ relativeTimeRange:
+ from: 1800
+ to: 0
+ datasourceUid: prometheus
+ model:
+ datasource:
+ type: prometheus
+ uid: prometheus
+ editorMode: code
+ exemplar: true
+ expr: ds_unb2_fpga_pol_rxgxb_temp{host="localhost",family="UNB2",member="H0"}
+ instant: false
+ interval: ""
+ intervalMs: 15000
+ legendFormat: TrRx board {{x}} node {{y}}
+ maxDataPoints: 43200
+ range: true
+ refId: C
+ - refId: G
+ relativeTimeRange:
+ from: 1800
+ to: 0
+ datasourceUid: __expr__
+ model:
+ conditions:
+ - evaluator:
+ params: []
+ type: gt
+ operator:
+ type: and
+ query:
+ params:
+ - G
+ reducer:
+ params: []
+ type: last
+ type: query
+ datasource:
+ type: __expr__
+ uid: __expr__
+ expression: C
+ intervalMs: 1000
+ maxDataPoints: 43200
+ reducer: max
+ refId: G
+ settings:
+ mode: dropNN
+ type: reduce
+ - refId: H
+ relativeTimeRange:
+ from: 1800
+ to: 0
+ datasourceUid: __expr__
+ model:
+ conditions:
+ - evaluator:
+ params:
+ - 85
+ type: gt
+ operator:
+ type: and
+ query:
+ params:
+ - H
+ reducer:
+ params: []
+ type: last
+ type: query
+ datasource:
+ type: __expr__
+ uid: __expr__
+ expression: G
+ intervalMs: 1000
+ maxDataPoints: 43200
+ refId: H
+ type: threshold
+ dashboardUid: c54bff42-4ff7-4e1d-8bf0-1f1c9cc0ea35
+ panelId: 4
+ noDataState: NoData
+ execErrState: Error
+ for: 5m
+ annotations:
+ __dashboardUid__: c54bff42-4ff7-4e1d-8bf0-1f1c9cc0ea35
+ __panelId__: "4"
+ description: FPGA POL rxgxb temperature for H0 has been above alarm conditions for 5 minutes.
+ summary: FPGA POL rxgxb temperature for H0 has been above alarm conditions for 5 minutes.
+ labels: {}
+ isPaused: false
+ - uid: fdloojx11up6oe
+ title: Uniboard2 FPGA POL rxgxb Temperatures L0
+ condition: H
+ data:
+ - refId: C
+ relativeTimeRange:
+ from: 1800
+ to: 0
+ datasourceUid: prometheus
+ model:
+ datasource:
+ type: prometheus
+ uid: prometheus
+ editorMode: code
+ exemplar: true
+ expr: ds_unb2_fpga_pol_rxgxb_temp{host="localhost",family="UNB2",member="L0"}
+ instant: false
+ interval: ""
+ intervalMs: 15000
+ legendFormat: TrRx board {{x}} node {{y}}
+ maxDataPoints: 43200
+ range: true
+ refId: C
+ - refId: G
+ relativeTimeRange:
+ from: 1800
+ to: 0
+ datasourceUid: __expr__
+ model:
+ conditions:
+ - evaluator:
+ params: []
+ type: gt
+ operator:
+ type: and
+ query:
+ params:
+ - G
+ reducer:
+ params: []
+ type: last
+ type: query
+ datasource:
+ type: __expr__
+ uid: __expr__
+ expression: C
+ intervalMs: 1000
+ maxDataPoints: 43200
+ reducer: max
+ refId: G
+ settings:
+ mode: dropNN
+ type: reduce
+ - refId: H
+ relativeTimeRange:
+ from: 1800
+ to: 0
+ datasourceUid: __expr__
+ model:
+ conditions:
+ - evaluator:
+ params:
+ - 85
+ type: gt
+ operator:
+ type: and
+ query:
+ params:
+ - H
+ reducer:
+ params: []
+ type: last
+ type: query
+ datasource:
+ type: __expr__
+ uid: __expr__
+ expression: G
+ intervalMs: 1000
+ maxDataPoints: 43200
+ refId: H
+ type: threshold
+ dashboardUid: c54bff42-4ff7-4e1d-8bf0-1f1c9cc0ea35
+ panelId: 4
+ noDataState: NoData
+ execErrState: Error
+ for: 5m
+ annotations:
+ __dashboardUid__: c54bff42-4ff7-4e1d-8bf0-1f1c9cc0ea35
+ __panelId__: "4"
+ description: FPGA POL rxgxb temperature for L0 has been above alarm conditions for 5 minutes.
+ summary: FPGA POL rxgxb temperature for L0 has been above alarm conditions for 5 minutes.
+ labels: {}
+ isPaused: false
+ - uid: cdlop06k1064gf
+ title: Uniboard2 FPGA POL rxgxb Temperatures L1
+ condition: H
+ data:
+ - refId: C
+ relativeTimeRange:
+ from: 1800
+ to: 0
+ datasourceUid: prometheus
+ model:
+ datasource:
+ type: prometheus
+ uid: prometheus
+ editorMode: code
+ exemplar: true
+ expr: ds_unb2_fpga_pol_rxgxb_temp{host="localhost",family="UNB2",member="L1"}
+ instant: false
+ interval: ""
+ intervalMs: 15000
+ legendFormat: TrRx board {{x}} node {{y}}
+ maxDataPoints: 43200
+ range: true
+ refId: C
+ - refId: G
+ relativeTimeRange:
+ from: 1800
+ to: 0
+ datasourceUid: __expr__
+ model:
+ conditions:
+ - evaluator:
+ params: []
+ type: gt
+ operator:
+ type: and
+ query:
+ params:
+ - G
+ reducer:
+ params: []
+ type: last
+ type: query
+ datasource:
+ type: __expr__
+ uid: __expr__
+ expression: C
+ intervalMs: 1000
+ maxDataPoints: 43200
+ reducer: max
+ refId: G
+ settings:
+ mode: dropNN
+ type: reduce
+ - refId: H
+ relativeTimeRange:
+ from: 1800
+ to: 0
+ datasourceUid: __expr__
+ model:
+ conditions:
+ - evaluator:
+ params:
+ - 85
+ type: gt
+ operator:
+ type: and
+ query:
+ params:
+ - H
+ reducer:
+ params: []
+ type: last
+ type: query
+ datasource:
+ type: __expr__
+ uid: __expr__
+ expression: G
+ intervalMs: 1000
+ maxDataPoints: 43200
+ refId: H
+ type: threshold
+ dashboardUid: c54bff42-4ff7-4e1d-8bf0-1f1c9cc0ea35
+ panelId: 4
+ noDataState: NoData
+ execErrState: Error
+ for: 5m
+ annotations:
+ __dashboardUid__: c54bff42-4ff7-4e1d-8bf0-1f1c9cc0ea35
+ __panelId__: "4"
+ description: FPGA POL rxgxb temperature for L1 has been above alarm conditions for 5 minutes.
+ summary: FPGA POL rxgxb temperature for L1 has been above alarm conditions for 5 minutes.
+ labels: {}
+ isPaused: false
+ - uid: bdlop7ejj1o8wa
+ title: Uniboard2 FPGA POL pgm Temperatures H0
+ condition: H
+ data:
+ - refId: E
+ relativeTimeRange:
+ from: 1800
+ to: 0
+ datasourceUid: prometheus
+ model:
+ datasource:
+ type: prometheus
+ uid: prometheus
+ editorMode: code
+ exemplar: true
+ expr: ds_unb2_fpga_pol_pgm_temp{host="localhost",family="UNB2",member="H0"}
+ interval: ""
+ intervalMs: 15000
+ legendFormat: IO board {{x}} node {{y}}
+ maxDataPoints: 43200
+ range: true
+ refId: E
+ - refId: G
+ relativeTimeRange:
+ from: 600
+ to: 0
+ datasourceUid: __expr__
+ model:
+ conditions:
+ - evaluator:
+ params: []
+ type: gt
+ operator:
+ type: and
+ query:
+ params:
+ - G
+ reducer:
+ params: []
+ type: last
+ type: query
+ datasource:
+ type: __expr__
+ uid: __expr__
+ expression: E
+ intervalMs: 1000
+ maxDataPoints: 43200
+ reducer: max
+ refId: G
+ settings:
+ mode: dropNN
+ type: reduce
+ - refId: H
+ relativeTimeRange:
+ from: 600
+ to: 0
+ datasourceUid: __expr__
+ model:
+ conditions:
+ - evaluator:
+ params:
+ - 85
+ type: gt
+ operator:
+ type: and
+ query:
+ params:
+ - H
+ reducer:
+ params: []
+ type: last
+ type: query
+ datasource:
+ type: __expr__
+ uid: __expr__
+ expression: G
+ intervalMs: 1000
+ maxDataPoints: 43200
+ refId: H
+ type: threshold
+ dashboardUid: c54bff42-4ff7-4e1d-8bf0-1f1c9cc0ea35
+ panelId: 4
+ noDataState: NoData
+ execErrState: Error
+ for: 5m
+ annotations:
+ __dashboardUid__: c54bff42-4ff7-4e1d-8bf0-1f1c9cc0ea35
+ __panelId__: "4"
+ description: FPGA POL pgm temperature for H0 has been above alarm conditions for 5 minutes.
+ summary: FPGA POL pgm temperature for H0 has been above alarm conditions for 5 minutes.
+ labels: {}
+ isPaused: false
+ - uid: edlopfjxrp43kb
+ title: Uniboard2 FPGA POL pgm Temperatures L0
+ condition: H
+ data:
+ - refId: E
+ relativeTimeRange:
+ from: 1800
+ to: 0
+ datasourceUid: prometheus
+ model:
+ datasource:
+ type: prometheus
+ uid: prometheus
+ editorMode: code
+ exemplar: true
+ expr: ds_unb2_fpga_pol_pgm_temp{host="localhost",family="UNB2",member="L0"}
+ instant: false
+ interval: ""
+ intervalMs: 15000
+ legendFormat: IO board {{x}} node {{y}}
+ maxDataPoints: 43200
+ range: true
+ refId: E
+ - refId: G
+ relativeTimeRange:
+ from: 1800
+ to: 0
+ datasourceUid: __expr__
+ model:
+ conditions:
+ - evaluator:
+ params: []
+ type: gt
+ operator:
+ type: and
+ query:
+ params:
+ - G
+ reducer:
+ params: []
+ type: last
+ type: query
+ datasource:
+ type: __expr__
+ uid: __expr__
+ expression: E
+ intervalMs: 1000
+ maxDataPoints: 43200
+ reducer: max
+ refId: G
+ settings:
+ mode: dropNN
+ type: reduce
+ - refId: H
+ relativeTimeRange:
+ from: 1800
+ to: 0
+ datasourceUid: __expr__
+ model:
+ conditions:
+ - evaluator:
+ params:
+ - 85
+ type: gt
+ operator:
+ type: and
+ query:
+ params:
+ - H
+ reducer:
+ params: []
+ type: last
+ type: query
+ datasource:
+ type: __expr__
+ uid: __expr__
+ expression: G
+ intervalMs: 1000
+ maxDataPoints: 43200
+ refId: H
+ type: threshold
+ dashboardUid: c54bff42-4ff7-4e1d-8bf0-1f1c9cc0ea35
+ panelId: 4
+ noDataState: NoData
+ execErrState: Error
+ for: 5m
+ annotations:
+ __dashboardUid__: c54bff42-4ff7-4e1d-8bf0-1f1c9cc0ea35
+ __panelId__: "4"
+ description: FPGA POL pgm temperature for L0 has been above alarm conditions for 5 minutes.
+ summary: FPGA POL pgm temperature for L0 has been above alarm conditions for 5 minutes.
+ labels: {}
+ isPaused: false
+ - uid: edlopgzduetq8d
+ title: Uniboard2 FPGA POL pgm Temperatures L1
+ condition: H
+ data:
+ - refId: E
+ relativeTimeRange:
+ from: 1800
+ to: 0
+ datasourceUid: prometheus
+ model:
+ datasource:
+ type: prometheus
+ uid: prometheus
+ editorMode: code
+ exemplar: true
+ expr: ds_unb2_fpga_pol_pgm_temp{host="localhost",family="UNB2",member="L1"}
+ instant: false
+ interval: ""
+ intervalMs: 15000
+ legendFormat: IO board {{x}} node {{y}}
+ maxDataPoints: 43200
+ range: true
+ refId: E
+ - refId: G
+ relativeTimeRange:
+ from: 1800
+ to: 0
+ datasourceUid: __expr__
+ model:
+ conditions:
+ - evaluator:
+ params: []
+ type: gt
+ operator:
+ type: and
+ query:
+ params:
+ - G
+ reducer:
+ params: []
+ type: last
+ type: query
+ datasource:
+ type: __expr__
+ uid: __expr__
+ expression: E
+ intervalMs: 1000
+ maxDataPoints: 43200
+ reducer: max
+ refId: G
+ settings:
+ mode: dropNN
+ type: reduce
+ - refId: H
+ relativeTimeRange:
+ from: 1800
+ to: 0
+ datasourceUid: __expr__
+ model:
+ conditions:
+ - evaluator:
+ params:
+ - 85
+ type: gt
+ operator:
+ type: and
+ query:
+ params:
+ - H
+ reducer:
+ params: []
+ type: last
+ type: query
+ datasource:
+ type: __expr__
+ uid: __expr__
+ expression: G
+ intervalMs: 1000
+ maxDataPoints: 43200
+ refId: H
+ type: threshold
+ dashboardUid: c54bff42-4ff7-4e1d-8bf0-1f1c9cc0ea35
+ panelId: 4
+ noDataState: NoData
+ execErrState: Error
+ for: 5m
+ annotations:
+ __dashboardUid__: c54bff42-4ff7-4e1d-8bf0-1f1c9cc0ea35
+ __panelId__: "4"
+ description: FPGA POL pgm temperature for L1 has been above alarm conditions for 5 minutes.
+ summary: FPGA POL pgm temperature for L1 has been above alarm conditions for 5 minutes.
+ labels: {}
+ isPaused: false
+ - uid: bdloppfezhxq8e
+ title: Uniboard2 FPGA POL hxgb Temperatures H0
+ condition: H
+ data:
+ - refId: F
+ relativeTimeRange:
+ from: 1800
+ to: 0
+ datasourceUid: prometheus
+ model:
+ datasource:
+ type: prometheus
+ uid: prometheus
+ editorMode: code
+ exemplar: true
+ expr: 'ds_unb2_fpga_pol_hgxb_temp{host="localhost",family="UNB2",member="H0"} '
+ interval: ""
+ intervalMs: 15000
+ legendFormat: HGXB board {{x}} node {{y}}
+ maxDataPoints: 43200
+ range: true
+ refId: F
+ - refId: G
+ relativeTimeRange:
+ from: 600
+ to: 0
+ datasourceUid: __expr__
+ model:
+ conditions:
+ - evaluator:
+ params: []
+ type: gt
+ operator:
+ type: and
+ query:
+ params:
+ - G
+ reducer:
+ params: []
+ type: last
+ type: query
+ datasource:
+ type: __expr__
+ uid: __expr__
+ expression: F
+ intervalMs: 1000
+ maxDataPoints: 43200
+ reducer: max
+ refId: G
+ settings:
+ mode: dropNN
+ type: reduce
+ - refId: H
+ relativeTimeRange:
+ from: 600
+ to: 0
+ datasourceUid: __expr__
+ model:
+ conditions:
+ - evaluator:
+ params:
+ - 85
+ type: gt
+ operator:
+ type: and
+ query:
+ params:
+ - H
+ reducer:
+ params: []
+ type: last
+ type: query
+ datasource:
+ type: __expr__
+ uid: __expr__
+ expression: G
+ intervalMs: 1000
+ maxDataPoints: 43200
+ refId: H
+ type: threshold
+ dashboardUid: c54bff42-4ff7-4e1d-8bf0-1f1c9cc0ea35
+ panelId: 4
+ noDataState: NoData
+ execErrState: Error
+ for: 5m
+ annotations:
+ __dashboardUid__: c54bff42-4ff7-4e1d-8bf0-1f1c9cc0ea35
+ __panelId__: "4"
+ description: FPGA POL Core temperature for H0 has been above alarm conditions for 5 minutes.
+ summary: FPGA POL Core temperature for H0 has been above alarm conditions for 5 minutes.
+ labels: {}
+ isPaused: false
+ - uid: bdlopvscihgjkd
+ title: Uniboard2 FPGA POL hxgb Temperatures L0
+ condition: H
+ data:
+ - refId: F
+ relativeTimeRange:
+ from: 1800
+ to: 0
+ datasourceUid: prometheus
+ model:
+ datasource:
+ type: prometheus
+ uid: prometheus
+ editorMode: code
+ exemplar: true
+ expr: 'ds_unb2_fpga_pol_hgxb_temp{host="localhost",family="UNB2",member="L0"} '
+ instant: false
+ interval: ""
+ intervalMs: 15000
+ legendFormat: HGXB board {{x}} node {{y}}
+ maxDataPoints: 43200
+ range: true
+ refId: F
+ - refId: G
+ relativeTimeRange:
+ from: 1800
+ to: 0
+ datasourceUid: __expr__
+ model:
+ conditions:
+ - evaluator:
+ params: []
+ type: gt
+ operator:
+ type: and
+ query:
+ params:
+ - G
+ reducer:
+ params: []
+ type: last
+ type: query
+ datasource:
+ type: __expr__
+ uid: __expr__
+ expression: F
+ intervalMs: 1000
+ maxDataPoints: 43200
+ reducer: max
+ refId: G
+ settings:
+ mode: dropNN
+ type: reduce
+ - refId: H
+ relativeTimeRange:
+ from: 1800
+ to: 0
+ datasourceUid: __expr__
+ model:
+ conditions:
+ - evaluator:
+ params:
+ - 85
+ type: gt
+ operator:
+ type: and
+ query:
+ params:
+ - H
+ reducer:
+ params: []
+ type: last
+ type: query
+ datasource:
+ type: __expr__
+ uid: __expr__
+ expression: G
+ intervalMs: 1000
+ maxDataPoints: 43200
+ refId: H
+ type: threshold
+ dashboardUid: c54bff42-4ff7-4e1d-8bf0-1f1c9cc0ea35
+ panelId: 4
+ noDataState: NoData
+ execErrState: Error
+ for: 5m
+ annotations:
+ __dashboardUid__: c54bff42-4ff7-4e1d-8bf0-1f1c9cc0ea35
+ __panelId__: "4"
+ description: FPGA POL Core temperature for L0 has been above alarm conditions for 5 minutes.
+ summary: FPGA POL Core temperature for L0 has been above alarm conditions for 5 minutes.
+ labels: {}
+ isPaused: false
+ - uid: fdlopwu07dkhsb
+ title: Uniboard2 FPGA POL hxgb Temperatures L1
+ condition: H
+ data:
+ - refId: F
+ relativeTimeRange:
+ from: 1800
+ to: 0
+ datasourceUid: prometheus
+ model:
+ datasource:
+ type: prometheus
+ uid: prometheus
+ editorMode: code
+ exemplar: true
+ expr: 'ds_unb2_fpga_pol_hgxb_temp{host="localhost",family="UNB2",member="L1"} '
+ instant: false
+ interval: ""
+ intervalMs: 15000
+ legendFormat: HGXB board {{x}} node {{y}}
+ maxDataPoints: 43200
+ range: true
+ refId: F
+ - refId: G
+ relativeTimeRange:
+ from: 1800
+ to: 0
+ datasourceUid: __expr__
+ model:
+ conditions:
+ - evaluator:
+ params: []
+ type: gt
+ operator:
+ type: and
+ query:
+ params:
+ - G
+ reducer:
+ params: []
+ type: last
+ type: query
+ datasource:
+ type: __expr__
+ uid: __expr__
+ expression: F
+ intervalMs: 1000
+ maxDataPoints: 43200
+ reducer: max
+ refId: G
+ settings:
+ mode: dropNN
+ type: reduce
+ - refId: H
+ relativeTimeRange:
+ from: 1800
+ to: 0
+ datasourceUid: __expr__
+ model:
+ conditions:
+ - evaluator:
+ params:
+ - 85
+ type: gt
+ operator:
+ type: and
+ query:
+ params:
+ - H
+ reducer:
+ params: []
+ type: last
+ type: query
+ datasource:
+ type: __expr__
+ uid: __expr__
+ expression: G
+ intervalMs: 1000
+ maxDataPoints: 43200
+ refId: H
+ type: threshold
+ dashboardUid: c54bff42-4ff7-4e1d-8bf0-1f1c9cc0ea35
+ panelId: 4
+ noDataState: NoData
+ execErrState: Error
+ for: 5m
+ annotations:
+ __dashboardUid__: c54bff42-4ff7-4e1d-8bf0-1f1c9cc0ea35
+ __panelId__: "4"
+ description: FPGA POL Core temperature for L1 has been above alarm conditions for 5 minutes.
+ summary: FPGA POL Core temperature for L1 has been above alarm conditions for 5 minutes.
+ labels: {}
+ isPaused: false
diff --git a/alerting/cp.yaml b/alerting/cp.yaml
new file mode 100644
index 0000000000000000000000000000000000000000..bb7e35ff11fb9efcfbb803602c8a94d2693aaa55
--- /dev/null
+++ b/alerting/cp.yaml
@@ -0,0 +1,12 @@
+apiVersion: 1
+contactPoints:
+ - orgId: 1
+ name: Corne Alarms
+ receivers:
+ - uid: slackalarms
+ type: slack
+ settings:
+ recipient: C04411Y8EAU # Corne slack user id
+ # Dummy token replaced by scripts
+ token: "xoxb-noop"
+ disableResolveMessage: true
diff --git a/alerting/policies.yaml b/alerting/policies.yaml
new file mode 100644
index 0000000000000000000000000000000000000000..de84c888cd86e5f15534a36fb613b0bfdb473132
--- /dev/null
+++ b/alerting/policies.yaml
@@ -0,0 +1,8 @@
+apiVersion: 1
+policies:
+ - orgId: 1
+ receiver: Corne Alarms
+ group_by:
+ - grafana_folder
+ group_interval: 1d
+ repeat_interval: 2d
diff --git a/dashboards/sensors.json b/dashboards/sensors.json
index 3e98d652865961b7477090507c04d5a7165a5070..040bd7f82e22ecd8d516c7501e2ff403f88387cb 100644
--- a/dashboards/sensors.json
+++ b/dashboards/sensors.json
@@ -3,7 +3,10 @@
"list": [
{
"builtIn": 1,
- "datasource": "-- Grafana --",
+ "datasource": {
+ "type": "datasource",
+ "uid": "grafana"
+ },
"enable": true,
"hide": true,
"iconColor": "rgba(0, 211, 255, 1)",
@@ -20,8 +23,8 @@
},
"editable": true,
"fiscalYearStartMonth": 0,
- "gnetId": null,
"graphTooltip": 0,
+ "id": 67,
"links": [
{
"asDropdown": true,
@@ -41,7 +44,10 @@
"liveNow": true,
"panels": [
{
- "datasource": null,
+ "datasource": {
+ "type": "prometheus",
+ "uid": "prometheus"
+ },
"gridPos": {
"h": 1,
"w": 24,
@@ -49,11 +55,23 @@
"y": 0
},
"id": 16,
+ "targets": [
+ {
+ "datasource": {
+ "type": "prometheus",
+ "uid": "prometheus"
+ },
+ "refId": "A"
+ }
+ ],
"title": "Temperatures",
"type": "row"
},
{
- "datasource": "Prometheus",
+ "datasource": {
+ "type": "prometheus",
+ "uid": "prometheus"
+ },
"description": "",
"fieldConfig": {
"defaults": {
@@ -61,6 +79,9 @@
"mode": "palette-classic"
},
"custom": {
+ "axisBorderShow": false,
+ "axisCenteredZero": false,
+ "axisColorMode": "text",
"axisLabel": "",
"axisPlacement": "auto",
"barAlignment": 0,
@@ -72,6 +93,7 @@
"tooltip": false,
"viz": false
},
+ "insertNulls": false,
"lineInterpolation": "linear",
"lineWidth": 1,
"pointSize": 5,
@@ -116,32 +138,39 @@
"options": {
"legend": {
"calcs": [],
- "displayMode": "hidden",
- "placement": "bottom"
+ "displayMode": "list",
+ "placement": "bottom",
+ "showLegend": false
},
"tooltip": {
- "mode": "single"
+ "mode": "single",
+ "sort": "none"
}
},
"pluginVersion": "8.1.2",
"targets": [
{
- "exemplar": true,
- "expr": "device_attribute{host=\"$station\",device=\"stat/sdpfirmware/$antennafield\",name=\"FPGA_temp_R\"} != 0",
- "format": "time_series",
+ "datasource": {
+ "type": "prometheus",
+ "uid": "prometheus"
+ },
+ "editorMode": "code",
+ "expr": "ds_fpga_temp{host=\"$station\",family=\"SDPFirmware\"} != 0",
"hide": false,
"instant": false,
- "interval": "",
"legendFormat": "{{x}}",
- "refId": "A"
+ "range": true,
+ "refId": "B"
}
],
"title": "FPGA Temperatures",
- "transformations": [],
"type": "timeseries"
},
{
- "datasource": "Prometheus",
+ "datasource": {
+ "type": "prometheus",
+ "uid": "prometheus"
+ },
"description": "",
"fieldConfig": {
"defaults": {
@@ -150,6 +179,9 @@
"seriesBy": "max"
},
"custom": {
+ "axisBorderShow": false,
+ "axisCenteredZero": false,
+ "axisColorMode": "text",
"axisLabel": "",
"axisPlacement": "auto",
"barAlignment": 0,
@@ -161,6 +193,7 @@
"tooltip": false,
"viz": false
},
+ "insertNulls": false,
"lineInterpolation": "linear",
"lineWidth": 1,
"pointSize": 5,
@@ -205,68 +238,108 @@
"options": {
"legend": {
"calcs": [],
- "displayMode": "hidden",
- "placement": "bottom"
+ "displayMode": "list",
+ "placement": "bottom",
+ "showLegend": false
},
"tooltip": {
- "mode": "single"
+ "mode": "single",
+ "sort": "none"
}
},
"targets": [
{
+ "datasource": {
+ "type": "prometheus",
+ "uid": "prometheus"
+ },
+ "editorMode": "code",
"exemplar": true,
- "expr": "device_attribute{host=\"$station\",device=\"stat/unb2/$subrack\",name=\"UNB2_FPGA_POL_CORE_TEMP_R\"} ",
+ "expr": "ds_unb2_fpga_pol_core_temp{host=\"$station\",family=\"UNB2\",member=\"$subrack\"} ",
"interval": "",
"legendFormat": "Core board {{x}} node {{y}}",
+ "range": true,
"refId": "A"
},
{
+ "datasource": {
+ "type": "prometheus",
+ "uid": "prometheus"
+ },
+ "editorMode": "code",
"exemplar": true,
- "expr": "device_attribute{host=\"$station\",device=\"stat/unb2/$subrack\",name=\"UNB2_FPGA_POL_ERAM_TEMP_R\"} ",
+ "expr": "ds_unb2_fpga_pol_eram_temp{host=\"$station\",family=\"UNB2\",member=\"$subrack\"}",
"hide": false,
"interval": "",
"legendFormat": "ERAM board {{x}} node {{y}}",
+ "range": true,
"refId": "B"
},
{
+ "datasource": {
+ "type": "prometheus",
+ "uid": "prometheus"
+ },
+ "editorMode": "code",
"exemplar": true,
- "expr": "device_attribute{host=\"$station\",device=\"stat/unb2/$subrack\",name=\"UNB2_FPGA_POL_RXGXB_TEMP_R\"} ",
+ "expr": "ds_unb2_fpga_pol_rxgxb_temp{host=\"$station\",family=\"UNB2\",member=\"$subrack\"}",
"hide": false,
"interval": "",
"legendFormat": "TrRx board {{x}} node {{y}}",
+ "range": true,
"refId": "C"
},
{
+ "datasource": {
+ "type": "prometheus",
+ "uid": "prometheus"
+ },
+ "editorMode": "code",
"exemplar": true,
- "expr": "device_attribute{host=\"$station\",device=\"stat/unb2/$subrack\",name=\"UNB2_FPGA_POL_TXGB_TEMP_R\"} ",
+ "expr": "ds_unb2_fpga_pol_txgb_temp{host=\"$station\",family=\"UNB2\",member=\"$subrack\"}",
"hide": false,
"interval": "",
"legendFormat": "TrHx board {{x}} node {{y}}",
+ "range": true,
"refId": "D"
},
{
+ "datasource": {
+ "type": "prometheus",
+ "uid": "prometheus"
+ },
+ "editorMode": "code",
"exemplar": true,
- "expr": "device_attribute{host=\"$station\",device=\"stat/unb2/$subrack\",name=\"UNB2_FPGA_POL_PGM_TEMP_R\"} ",
+ "expr": "ds_unb2_fpga_pol_pgm_temp{host=\"$station\",family=\"UNB2\",member=\"$subrack\"}",
"hide": false,
"interval": "",
"legendFormat": "IO board {{x}} node {{y}}",
+ "range": true,
"refId": "E"
},
{
+ "datasource": {
+ "type": "prometheus",
+ "uid": "prometheus"
+ },
+ "editorMode": "code",
"exemplar": true,
- "expr": "device_attribute{host=\"$station\",device=\"stat/unb2/$subrack\",name=\"UNB2_FPGA_POL_HGXB_TEMP_R\"} ",
+ "expr": "ds_unb2_fpga_pol_hgxb_temp{host=\"$station\",family=\"UNB2\",member=\"$subrack\"} ",
"hide": false,
"interval": "",
"legendFormat": "HGXB board {{x}} node {{y}}",
+ "range": true,
"refId": "F"
}
],
"title": "Uniboard2 FPGA POL Temperatures",
- "transformations": [],
"type": "timeseries"
},
{
- "datasource": "Prometheus",
+ "datasource": {
+ "type": "prometheus",
+ "uid": "prometheus"
+ },
"description": "",
"fieldConfig": {
"defaults": {
@@ -275,6 +348,9 @@
"seriesBy": "max"
},
"custom": {
+ "axisBorderShow": false,
+ "axisCenteredZero": false,
+ "axisColorMode": "text",
"axisLabel": "",
"axisPlacement": "auto",
"barAlignment": 0,
@@ -286,6 +362,7 @@
"tooltip": false,
"viz": false
},
+ "insertNulls": false,
"lineInterpolation": "linear",
"lineWidth": 1,
"pointSize": 5,
@@ -330,15 +407,21 @@
"options": {
"legend": {
"calcs": [],
- "displayMode": "hidden",
- "placement": "bottom"
+ "displayMode": "list",
+ "placement": "bottom",
+ "showLegend": false
},
"tooltip": {
- "mode": "single"
+ "mode": "single",
+ "sort": "none"
}
},
"targets": [
{
+ "datasource": {
+ "type": "prometheus",
+ "uid": "prometheus"
+ },
"exemplar": true,
"expr": "device_attribute{host=\"$station\",device=\"stat/unb2/$subrack\",name=\"UNB2_FPGA_QSFP_CAGE_TEMP_R\"}",
"interval": "",
@@ -350,7 +433,10 @@
"type": "timeseries"
},
{
- "datasource": "Prometheus",
+ "datasource": {
+ "type": "prometheus",
+ "uid": "prometheus"
+ },
"description": "",
"fieldConfig": {
"defaults": {
@@ -359,6 +445,9 @@
"seriesBy": "max"
},
"custom": {
+ "axisBorderShow": false,
+ "axisCenteredZero": false,
+ "axisColorMode": "text",
"axisLabel": "",
"axisPlacement": "auto",
"barAlignment": 0,
@@ -370,6 +459,7 @@
"tooltip": false,
"viz": false
},
+ "insertNulls": false,
"lineInterpolation": "linear",
"lineWidth": 1,
"pointSize": 5,
@@ -414,15 +504,21 @@
"options": {
"legend": {
"calcs": [],
- "displayMode": "hidden",
- "placement": "bottom"
+ "displayMode": "list",
+ "placement": "bottom",
+ "showLegend": false
},
"tooltip": {
- "mode": "single"
+ "mode": "single",
+ "sort": "none"
}
},
"targets": [
{
+ "datasource": {
+ "type": "prometheus",
+ "uid": "prometheus"
+ },
"exemplar": true,
"expr": "device_attribute{host=\"$station\",device=\"stat/unb2/$subrack\",name=\"UNB2_FPGA_DDR4_SLOT_TEMP_R\"}",
"interval": "",
@@ -434,7 +530,10 @@
"type": "timeseries"
},
{
- "datasource": "Prometheus",
+ "datasource": {
+ "type": "prometheus",
+ "uid": "prometheus"
+ },
"description": "",
"fieldConfig": {
"defaults": {
@@ -442,6 +541,9 @@
"mode": "palette-classic"
},
"custom": {
+ "axisBorderShow": false,
+ "axisCenteredZero": false,
+ "axisColorMode": "text",
"axisLabel": "",
"axisPlacement": "auto",
"barAlignment": 0,
@@ -453,6 +555,7 @@
"tooltip": false,
"viz": false
},
+ "insertNulls": false,
"lineInterpolation": "linear",
"lineWidth": 1,
"pointSize": 5,
@@ -497,16 +600,22 @@
"options": {
"legend": {
"calcs": [],
- "displayMode": "hidden",
- "placement": "bottom"
+ "displayMode": "list",
+ "placement": "bottom",
+ "showLegend": false
},
"tooltip": {
- "mode": "single"
+ "mode": "single",
+ "sort": "none"
}
},
"pluginVersion": "8.1.2",
"targets": [
{
+ "datasource": {
+ "type": "prometheus",
+ "uid": "prometheus"
+ },
"exemplar": true,
"expr": "device_attribute{host=\"$station\",device=\"$recv\",name=\"RCU_TEMP_R\"}",
"format": "time_series",
@@ -518,11 +627,13 @@
}
],
"title": "RCU Temperatures",
- "transformations": [],
"type": "timeseries"
},
{
- "datasource": "Prometheus",
+ "datasource": {
+ "type": "prometheus",
+ "uid": "prometheus"
+ },
"description": "Temperatures reported by APSCT and APSPU",
"fieldConfig": {
"defaults": {
@@ -531,6 +642,9 @@
"seriesBy": "max"
},
"custom": {
+ "axisBorderShow": false,
+ "axisCenteredZero": false,
+ "axisColorMode": "text",
"axisLabel": "",
"axisPlacement": "auto",
"barAlignment": 0,
@@ -542,6 +656,7 @@
"tooltip": false,
"viz": false
},
+ "insertNulls": false,
"lineInterpolation": "linear",
"lineWidth": 1,
"pointSize": 5,
@@ -586,15 +701,21 @@
"options": {
"legend": {
"calcs": [],
- "displayMode": "hidden",
- "placement": "bottom"
+ "displayMode": "list",
+ "placement": "bottom",
+ "showLegend": false
},
"tooltip": {
- "mode": "single"
+ "mode": "single",
+ "sort": "none"
}
},
"targets": [
{
+ "datasource": {
+ "type": "prometheus",
+ "uid": "prometheus"
+ },
"exemplar": true,
"expr": "device_attribute{host=\"$station\",device=\"stat/apsct/$subrack\",name=~\"APSCT_TEMP_R\"}",
"interval": "",
@@ -602,6 +723,10 @@
"refId": "A"
},
{
+ "datasource": {
+ "type": "prometheus",
+ "uid": "prometheus"
+ },
"exemplar": true,
"expr": "device_attribute{host=\"$station\",device=\"stat/apspu/$subrack\",name=~\"APSPU_.*_TEMP_R\"}",
"hide": false,
@@ -614,7 +739,10 @@
"type": "timeseries"
},
{
- "datasource": "Prometheus",
+ "datasource": {
+ "type": "prometheus",
+ "uid": "prometheus"
+ },
"description": "Temperature sensors of the power supply on each board",
"fieldConfig": {
"defaults": {
@@ -623,6 +751,9 @@
"seriesBy": "max"
},
"custom": {
+ "axisBorderShow": false,
+ "axisCenteredZero": false,
+ "axisColorMode": "text",
"axisLabel": "",
"axisPlacement": "auto",
"barAlignment": 0,
@@ -634,6 +765,7 @@
"tooltip": false,
"viz": false
},
+ "insertNulls": false,
"lineInterpolation": "linear",
"lineWidth": 1,
"pointSize": 5,
@@ -671,22 +803,28 @@
"gridPos": {
"h": 8,
"w": 5,
- "x": 15,
+ "x": 10,
"y": 9
},
"id": 8,
"options": {
"legend": {
"calcs": [],
- "displayMode": "hidden",
- "placement": "bottom"
+ "displayMode": "list",
+ "placement": "bottom",
+ "showLegend": false
},
"tooltip": {
- "mode": "single"
+ "mode": "single",
+ "sort": "none"
}
},
"targets": [
{
+ "datasource": {
+ "type": "prometheus",
+ "uid": "prometheus"
+ },
"exemplar": true,
"expr": "device_attribute{host=\"$station\",device=\"stat/unb2/$subrack\",name=\"UNB2_POL_QSFP_N01_TEMP_R\"} ",
"interval": "",
@@ -694,6 +832,10 @@
"refId": "A"
},
{
+ "datasource": {
+ "type": "prometheus",
+ "uid": "prometheus"
+ },
"exemplar": true,
"expr": "device_attribute{host=\"$station\",device=\"stat/unb2/$subrack\",name=\"UNB2_POL_QSFP_N23_TEMP_R\"} ",
"hide": false,
@@ -702,6 +844,10 @@
"refId": "B"
},
{
+ "datasource": {
+ "type": "prometheus",
+ "uid": "prometheus"
+ },
"exemplar": true,
"expr": "device_attribute{host=\"$station\",device=\"stat/unb2/$subrack\",name=\"UNB2_POL_SWITCH_1V2_TEMP_R\"} ",
"hide": false,
@@ -710,6 +856,10 @@
"refId": "C"
},
{
+ "datasource": {
+ "type": "prometheus",
+ "uid": "prometheus"
+ },
"exemplar": true,
"expr": "device_attribute{host=\"$station\",device=\"stat/unb2/$subrack\",name=\"UNB2_POL_SWITCH_PHY_TEMP_R\"} ",
"hide": false,
@@ -718,6 +868,10 @@
"refId": "D"
},
{
+ "datasource": {
+ "type": "prometheus",
+ "uid": "prometheus"
+ },
"exemplar": true,
"expr": "device_attribute{host=\"$station\",device=\"stat/unb2/$subrack\",name=\"UNB2_POL_CLOCK_TEMP_R\"} ",
"hide": false,
@@ -726,6 +880,10 @@
"refId": "E"
},
{
+ "datasource": {
+ "type": "prometheus",
+ "uid": "prometheus"
+ },
"exemplar": true,
"expr": "device_attribute{host=\"$station\",device=\"stat/unb2/$subrack\",name=\"UNB2_DC_DC_48V_12V_TEMP_R\"} ",
"hide": false,
@@ -739,7 +897,10 @@
},
{
"collapsed": true,
- "datasource": null,
+ "datasource": {
+ "type": "prometheus",
+ "uid": "prometheus"
+ },
"gridPos": {
"h": 1,
"w": 24,
@@ -748,11 +909,23 @@
},
"id": 18,
"panels": [],
+ "targets": [
+ {
+ "datasource": {
+ "type": "prometheus",
+ "uid": "prometheus"
+ },
+ "refId": "A"
+ }
+ ],
"title": "Voltages",
"type": "row"
},
{
- "datasource": "Prometheus",
+ "datasource": {
+ "type": "prometheus",
+ "uid": "prometheus"
+ },
"description": "Voltage sensors of the power supplies of the APS",
"fieldConfig": {
"defaults": {
@@ -761,6 +934,9 @@
"seriesBy": "max"
},
"custom": {
+ "axisBorderShow": false,
+ "axisCenteredZero": false,
+ "axisColorMode": "text",
"axisLabel": "",
"axisPlacement": "auto",
"axisSoftMax": 10,
@@ -774,6 +950,7 @@
"tooltip": false,
"viz": false
},
+ "insertNulls": false,
"lineInterpolation": "linear",
"lineWidth": 1,
"pointSize": 5,
@@ -818,15 +995,21 @@
"options": {
"legend": {
"calcs": [],
- "displayMode": "hidden",
- "placement": "bottom"
+ "displayMode": "list",
+ "placement": "bottom",
+ "showLegend": false
},
"tooltip": {
- "mode": "single"
+ "mode": "single",
+ "sort": "none"
}
},
"targets": [
{
+ "datasource": {
+ "type": "prometheus",
+ "uid": "prometheus"
+ },
"exemplar": true,
"expr": "device_attribute{host=\"$station\",device=\"stat/apspu/$subrack\",name=~\"APSPU_.*_VOUT_R\"}",
"interval": "",
@@ -838,7 +1021,10 @@
"type": "timeseries"
},
{
- "datasource": "Prometheus",
+ "datasource": {
+ "type": "prometheus",
+ "uid": "prometheus"
+ },
"description": "Voltage sensors of each node on each board",
"fieldConfig": {
"defaults": {
@@ -846,6 +1032,9 @@
"mode": "palette-classic"
},
"custom": {
+ "axisBorderShow": false,
+ "axisCenteredZero": false,
+ "axisColorMode": "text",
"axisLabel": "",
"axisPlacement": "auto",
"axisSoftMax": 2,
@@ -859,6 +1048,7 @@
"tooltip": false,
"viz": false
},
+ "insertNulls": false,
"lineInterpolation": "linear",
"lineWidth": 1,
"pointSize": 5,
@@ -903,15 +1093,21 @@
"options": {
"legend": {
"calcs": [],
- "displayMode": "hidden",
- "placement": "bottom"
+ "displayMode": "list",
+ "placement": "bottom",
+ "showLegend": false
},
"tooltip": {
- "mode": "single"
+ "mode": "single",
+ "sort": "none"
}
},
"targets": [
{
+ "datasource": {
+ "type": "prometheus",
+ "uid": "prometheus"
+ },
"exemplar": true,
"expr": "device_attribute{host=\"$station\",device=\"stat/unb2/$subrack\",name=\"UNB2_FPGA_POL_CORE_VOUT_R\"}",
"interval": "",
@@ -919,6 +1115,10 @@
"refId": "A"
},
{
+ "datasource": {
+ "type": "prometheus",
+ "uid": "prometheus"
+ },
"exemplar": true,
"expr": "device_attribute{host=\"$station\",device=\"stat/unb2/$subrack\",name=\"UNB2_FPGA_POL_ERAM_VOUT_R\"}",
"hide": false,
@@ -927,6 +1127,10 @@
"refId": "B"
},
{
+ "datasource": {
+ "type": "prometheus",
+ "uid": "prometheus"
+ },
"exemplar": true,
"expr": "device_attribute{host=\"$station\",device=\"stat/unb2/$subrack\",name=\"UNB2_FPGA_POL_RXGXB_VOUT_R\"}",
"hide": false,
@@ -935,6 +1139,10 @@
"refId": "C"
},
{
+ "datasource": {
+ "type": "prometheus",
+ "uid": "prometheus"
+ },
"exemplar": true,
"expr": "device_attribute{host=\"$station\",device=\"stat/unb2/$subrack\",name=\"UNB2_FPGA_POL_TXGB_VOUT_R\"}",
"hide": false,
@@ -943,6 +1151,10 @@
"refId": "D"
},
{
+ "datasource": {
+ "type": "prometheus",
+ "uid": "prometheus"
+ },
"exemplar": true,
"expr": "device_attribute{host=\"$station\",device=\"stat/unb2/$subrack\",name=\"UNB2_FPGA_POL_PGM_VOUT_R\"}",
"hide": false,
@@ -955,7 +1167,10 @@
"type": "timeseries"
},
{
- "datasource": "Prometheus",
+ "datasource": {
+ "type": "prometheus",
+ "uid": "prometheus"
+ },
"description": "Voltage sensors of the power supply on each board",
"fieldConfig": {
"defaults": {
@@ -964,6 +1179,9 @@
"seriesBy": "max"
},
"custom": {
+ "axisBorderShow": false,
+ "axisCenteredZero": false,
+ "axisColorMode": "text",
"axisLabel": "",
"axisPlacement": "auto",
"axisSoftMin": 0,
@@ -976,6 +1194,7 @@
"tooltip": false,
"viz": false
},
+ "insertNulls": false,
"lineInterpolation": "linear",
"lineWidth": 1,
"pointSize": 5,
@@ -1020,15 +1239,21 @@
"options": {
"legend": {
"calcs": [],
- "displayMode": "hidden",
- "placement": "bottom"
+ "displayMode": "list",
+ "placement": "bottom",
+ "showLegend": false
},
"tooltip": {
- "mode": "single"
+ "mode": "single",
+ "sort": "none"
}
},
"targets": [
{
+ "datasource": {
+ "type": "prometheus",
+ "uid": "prometheus"
+ },
"exemplar": true,
"expr": "device_attribute{host=\"$station\",device=\"stat/unb2/$subrack\",name=\"UNB2_POL_QSFP_N01_VOUT_R\"}",
"interval": "",
@@ -1036,6 +1261,10 @@
"refId": "A"
},
{
+ "datasource": {
+ "type": "prometheus",
+ "uid": "prometheus"
+ },
"exemplar": true,
"expr": "device_attribute{host=\"$station\",device=\"stat/unb2/$subrack\",name=\"UNB2_POL_QSFP_N23_VOUT_R\"}",
"hide": false,
@@ -1044,6 +1273,10 @@
"refId": "B"
},
{
+ "datasource": {
+ "type": "prometheus",
+ "uid": "prometheus"
+ },
"exemplar": true,
"expr": "device_attribute{host=\"$station\",device=\"stat/unb2/$subrack\",name=\"UNB2_POL_SWITCH_1V2_VOUT_R\"}",
"hide": false,
@@ -1052,6 +1285,10 @@
"refId": "C"
},
{
+ "datasource": {
+ "type": "prometheus",
+ "uid": "prometheus"
+ },
"exemplar": true,
"expr": "device_attribute{host=\"$station\",device=\"stat/unb2/$subrack\",name=\"UNB2_POL_SWITCH_PHY_VOUT_R\"}",
"hide": false,
@@ -1060,6 +1297,10 @@
"refId": "D"
},
{
+ "datasource": {
+ "type": "prometheus",
+ "uid": "prometheus"
+ },
"exemplar": true,
"expr": "device_attribute{host=\"$station\",device=\"stat/unb2/$subrack\",name=\"UNB2_POL_CLOCK_VOUT_R\"}",
"hide": false,
@@ -1068,6 +1309,10 @@
"refId": "E"
},
{
+ "datasource": {
+ "type": "prometheus",
+ "uid": "prometheus"
+ },
"exemplar": true,
"expr": "device_attribute{host=\"$station\",device=\"stat/unb2/$subrack\",name=\"UNB2_DC_DC_48V_12V_VOUT_R\"}",
"hide": false,
@@ -1081,7 +1326,10 @@
},
{
"collapsed": true,
- "datasource": null,
+ "datasource": {
+ "type": "prometheus",
+ "uid": "prometheus"
+ },
"gridPos": {
"h": 1,
"w": 24,
@@ -1090,11 +1338,23 @@
},
"id": 20,
"panels": [],
+ "targets": [
+ {
+ "datasource": {
+ "type": "prometheus",
+ "uid": "prometheus"
+ },
+ "refId": "A"
+ }
+ ],
"title": "Clock stability",
"type": "row"
},
{
- "datasource": "Prometheus",
+ "datasource": {
+ "type": "prometheus",
+ "uid": "prometheus"
+ },
"description": "Measured difference between PTP and PPS",
"fieldConfig": {
"defaults": {
@@ -1102,6 +1362,9 @@
"mode": "thresholds"
},
"custom": {
+ "axisBorderShow": false,
+ "axisCenteredZero": false,
+ "axisColorMode": "text",
"axisLabel": "",
"axisPlacement": "auto",
"barAlignment": 0,
@@ -1113,6 +1376,7 @@
"tooltip": false,
"viz": false
},
+ "insertNulls": false,
"lineInterpolation": "linear",
"lineWidth": 1,
"pointSize": 5,
@@ -1135,8 +1399,7 @@
"mode": "absolute",
"steps": [
{
- "color": "red",
- "value": null
+ "color": "red"
},
{
"color": "green",
@@ -1162,16 +1425,22 @@
"options": {
"legend": {
"calcs": [],
- "displayMode": "hidden",
- "placement": "bottom"
+ "displayMode": "list",
+ "placement": "bottom",
+ "showLegend": false
},
"tooltip": {
- "mode": "single"
+ "mode": "single",
+ "sort": "none"
}
},
"pluginVersion": "8.1.2",
"targets": [
{
+ "datasource": {
+ "type": "prometheus",
+ "uid": "prometheus"
+ },
"exemplar": true,
"expr": "device_attribute{host=\"$station\",device=\"stat/sdpfirmware/$antennafield\",name=\"TR_tod_pps_delta_R\"}",
"format": "time_series",
@@ -1183,12 +1452,11 @@
}
],
"title": "FPGA Clock offset",
- "transformations": [],
"type": "timeseries"
}
],
- "schemaVersion": 31,
- "style": "dark",
+ "refresh": "",
+ "schemaVersion": 39,
"tags": [
"station"
],
@@ -1197,14 +1465,14 @@
{
"current": {
"selected": false,
- "text": "test-lcu2",
- "value": "test-lcu2"
+ "text": "localhost",
+ "value": "localhost"
},
"datasource": {
"type": "prometheus",
"uid": "prometheus"
},
- "definition": "scrape_series_added\n",
+ "definition": "label_values(ds_state,host)",
"description": "",
"hide": 0,
"includeAll": false,
@@ -1213,108 +1481,73 @@
"name": "station",
"options": [],
"query": {
- "query": "scrape_series_added\n",
- "refId": "StandardVariableQuery"
- },
- "refresh": 1,
- "regex": "/.*host=\"(?<text>[^\"]*)\".*/",
- "skipUrlSync": false,
- "sort": 1,
- "type": "query"
- },
- {
- "current": {},
- "definition": "label_values(device_attribute{device=\"stat/stationmanager/1\", host=\"$station\", name=\"station_name_R\"},str_value)",
- "description": "Station Name retrieved from device-stationmanager",
- "hide": 0,
- "includeAll": false,
- "label": "Station Name",
- "multi": false,
- "name": "station_name",
- "options": [],
- "query": {
- "query": "label_values(device_attribute{device=\"stat/stationmanager/1\", host=\"$station\", name=\"station_name_R\"},str_value)",
+ "qryType": 1,
+ "query": "label_values(ds_state,host)",
"refId": "PrometheusVariableQueryEditor-VariableQuery"
},
"refresh": 1,
"regex": "",
"skipUrlSync": false,
- "sort": 0,
+ "sort": 1,
"type": "query"
},
{
"current": {
- "selected": true,
- "text": "h0",
- "value": "h0"
+ "selected": false,
+ "text": "LBA",
+ "value": "LBA"
},
"datasource": {
"type": "prometheus",
"uid": "prometheus"
},
- "definition": "device_attribute{host=\"$station\", device=~\"stat/apspu/.*\", name=\"State\"}",
+ "definition": "label_values(ds_state{family=~\"AFL|AFH\", host=\"$station\"},member)",
"description": "",
"hide": 0,
"includeAll": false,
- "label": "Subrack",
+ "label": "Antennafield",
"multi": false,
- "name": "subrack",
+ "name": "antennafield",
"options": [],
"query": {
- "query": "device_attribute{host=\"$station\", device=~\"stat/apspu/.*\", name=\"State\"}",
- "refId": "StandardVariableQuery"
+ "qryType": 1,
+ "query": "label_values(ds_state{family=~\"AFL|AFH\", host=\"$station\"},member)",
+ "refId": "PrometheusVariableQueryEditor-VariableQuery"
},
"refresh": 1,
- "regex": "/.*device=\"stat/apspu/(?<text>[^\"]*)\".*/",
+ "regex": "",
"skipUrlSync": false,
"sort": 1,
"type": "query"
},
{
- "current": {},
+ "current": {
+ "selected": false,
+ "text": "H0",
+ "value": "H0"
+ },
"datasource": {
"type": "prometheus",
"uid": "prometheus"
},
- "definition": "device_attribute{host=\"$station\", device=~\"stat/af[hl]/.*\", name=\"State\"}",
+ "definition": "label_values(ds_state{family=\"APS\", host=\"$station\"},member)",
"description": "",
"hide": 0,
"includeAll": false,
- "label": "Antennafield",
+ "label": "Subrack",
"multi": false,
- "name": "antennafield",
+ "name": "subrack",
"options": [],
"query": {
- "query": "device_attribute{host=\"$station\", device=~\"stat/af[hl]/.*\", name=\"State\"}",
- "refId": "StandardVariableQuery"
+ "qryType": 1,
+ "query": "label_values(ds_state{family=\"APS\", host=\"$station\"},member)",
+ "refId": "PrometheusVariableQueryEditor-VariableQuery"
},
"refresh": 1,
- "regex": "/.*device=\"stat/af[hl]/(?<text>[^\"]*)\".*/",
+ "regex": "",
"skipUrlSync": false,
"sort": 1,
"type": "query"
- },
- {
- "current": {
- "selected": false,
- "text": "stat/recv/h0",
- "value": "stat/recv/h0"
- },
- "definition": "device_attribute{host=\"$station\", device=~\"stat/(recv|rcu2l|rcu2h|recvh|recvl)/$subrack\", name=\"State\"}",
- "hide": 2,
- "includeAll": false,
- "multi": false,
- "name": "recv",
- "options": [],
- "query": {
- "query": "device_attribute{host=\"$station\", device=~\"stat/(recv|rcu2l|rcu2h|recvh|recvl)/$subrack\", name=\"State\"}",
- "refId": "StandardVariableQuery"
- },
- "refresh": 1,
- "regex": "/.*device=\"(?<text>[^\"]*)\".*/",
- "skipUrlSync": false,
- "sort": 0,
- "type": "query"
}
]
},
@@ -1325,6 +1558,7 @@
"timepicker": {},
"timezone": "",
"title": "Sensors",
- "uid": "KMRmQzd7z",
- "version": 1
-}
+ "uid": "c54bff42-4ff7-4e1d-8bf0-1f1c9cc0ea35",
+ "version": 11,
+ "weekStart": ""
+}
\ No newline at end of file
diff --git a/datasources/grafanaapi.yaml b/datasources/grafanaapi.yaml
deleted file mode 100644
index a2310cdf2f4432c09581b1f60bbf9ec16a573606..0000000000000000000000000000000000000000
--- a/datasources/grafanaapi.yaml
+++ /dev/null
@@ -1,36 +0,0 @@
-apiVersion: 1
-
-datasources:
- # <string, required> name of the datasource. Required
- - name: Grafana API
- # <string, required> datasource type. Required
- type: yesoreyeram-infinity-datasource
- # <string, required> access mode. proxy or direct (Server or Browser in the UI). Required
- access: proxy
- # <int> org id. will default to orgId 1 if not specified
- orgId: 1
- # <string> custom UID which can be used to reference this datasource in other parts of the configuration, if not specified will be generated automatically
- uid: grafanaapi
- # <string> url
- url: http://localhost:3000/api
- # <string> Deprecated, use secureJsonData.password
- password:
- # <string> database user, if used
- user: postgres
- # <string> database name, if used
- database: hdb
- # <bool> enable/disable basic auth
- basicAuth: false
- # <string> basic auth username
- basicAuthUser:
- # <string> Deprecated, use secureJsonData.basicAuthPassword
- basicAuthPassword:
- # <bool> enable/disable with credentials headers
- withCredentials:
- # <bool> mark as default datasource. Max one per org
- isDefault: false
- # <map> fields that will be converted to json and stored in jsonData
- version: 1
- # <bool> allow users to edit datasources from the UI.
- editable: false
-
diff --git a/datasources/loki.yaml b/datasources/loki.yaml
index f9108f15f3791de72fb8c80cc24ae156e0bfea73..d98b2971fe70b60aaa274f7c5602cd8076fda70f 100644
--- a/datasources/loki.yaml
+++ b/datasources/loki.yaml
@@ -13,21 +13,6 @@ datasources:
uid: loki
# <string> url
url: http://loki:3100
- # <string> Deprecated, use secureJsonData.password
- password:
- # <string> database user, if used
- user:
- # <string> database name, if used
- database:
- # <bool> enable/disable basic auth
- basicAuth: false
- # <string> basic auth username
- basicAuthUser:
- # <string> Deprecated, use secureJsonData.basicAuthPassword
- basicAuthPassword:
- # <bool> enable/disable with credentials headers
- withCredentials:
- # <bool> mark as default datasource. Max one per org
isDefault: false
# <map> fields that will be converted to json and stored in jsonData
jsonData:
@@ -37,9 +22,3 @@ datasources:
logMessageField:
maxConcurrentShardRequests: 5
timeField: "@timestamp"
- # <string> json object of data that will be encrypted.
- secureJsonData:
- version: 1
- # <bool> allow users to edit datasources from the UI.
- editable: false
-
diff --git a/datasources/prometheus.yaml b/datasources/prometheus.yaml
index efea8bd474db6eb6c9865c1731be5e4a46c42fcc..eeea0044d73713575ffcec2e8b234210d58f5902 100644
--- a/datasources/prometheus.yaml
+++ b/datasources/prometheus.yaml
@@ -13,27 +13,4 @@ datasources:
uid: prometheus
# <string> url
url: http://prometheus:9090
- # <string> Deprecated, use secureJsonData.password
- password:
- # <string> database user, if used
- user:
- # <string> database name, if used
- database:
- # <bool> enable/disable basic auth
- basicAuth: false
- # <string> basic auth username
- basicAuthUser:
- # <string> Deprecated, use secureJsonData.basicAuthPassword
- basicAuthPassword:
- # <bool> enable/disable with credentials headers
- withCredentials:
- # <bool> mark as default datasource. Max one per org
isDefault: true
- # <map> fields that will be converted to json and stored in jsonData
- jsonData:
- httpMethod: POST
- # <string> json object of data that will be encrypted.
- secureJsonData:
- version: 1
- # <bool> allow users to edit datasources from the UI.
- editable: false
diff --git a/import-rules.sh b/import-rules.sh
deleted file mode 100755
index b3d5871bd532387a5386b8fdfee0331c0daeddc2..0000000000000000000000000000000000000000
--- a/import-rules.sh
+++ /dev/null
@@ -1,24 +0,0 @@
-#! /bin/bash
-
-until curl -s -X GET "http://localhost:3000" -o /dev/null
-do
- echo "Wait until grafana is ready..."
- sleep 5
-done
-
-# Create API key
-API_RESPONSE=`curl -s -X POST -H "Content-Type: application/json" -d '{"name":"apikeycurl", "role": "Admin"}' http://admin:admin@localhost:3000/api/auth/keys`
-API_KEY=`echo "$API_RESPONSE" | jq -j '.key'`
-API_KEY_ID=`echo "$API_RESPONSE" | jq -j '.id'`
-
-# Import alerts
-curl -X POST http://localhost:3000/api/alertmanager/grafana/config/api/v1/alerts -H 'Content-Type: application/json' -H 'Accept: application/json' -H "Authorization: Bearer $API_KEY" -d '@alerting.json'
-curl -X POST http://localhost:3000/api/folders -H 'Content-Type: application/json' -H 'Accept: application/json' -H "Authorization: Bearer $API_KEY" -d '{"title": "station"}'
-
-# Import station rules
-jq -c '.station[]' rules.json | while read rule; do
- echo $rule | curl -X POST http://localhost:3000/api/ruler/grafana/api/v1/rules/station -H 'Content-Type: application/json' -H 'Accept: application/json' -H "Authorization: Bearer $API_KEY" -d '@-'
-done
-
-# Cleanup api key
-curl -s -X DELETE "http://localhost:3000/api/auth/keys/$API_KEY_ID" -H 'Content-Type: application/json' -H 'Accept: application/json' -H "Authorization: Bearer $API_KEY"
\ No newline at end of file
diff --git a/imports/README.md b/imports/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..fa913d1319a86a0542e21390fe03ec1a477aff87
--- /dev/null
+++ b/imports/README.md
@@ -0,0 +1,4 @@
+# Imports
+
+These files are manually copied by the Dockerfile, intended for Alert but not
+compatible with alert provisioning (disables admin API)
\ No newline at end of file
diff --git a/imports/populate-tokens.sh b/imports/populate-tokens.sh
new file mode 100755
index 0000000000000000000000000000000000000000..0a350a11c1e93443e840cdb12333f6b15a8499cc
--- /dev/null
+++ b/imports/populate-tokens.sh
@@ -0,0 +1,10 @@
+#! /bin/bash
+# Populate tokens from environment
+
+ALERT_DIR="/etc/grafana/provisioning/alerting/"
+
+if [[ -n "${SLACK_TOKEN}" ]]; then
+ yq -i '(.contactPoints[].receivers[] | select(has("type")) | select(.type == "slack")) .settings.token = ("${SLACK_TOKEN}" | envsubst)' ${ALERT_DIR}/cp.yaml
+else
+ echo "SLACK_TOKEN not set alarm notifications for slack will not work!" >&2
+fi
diff --git a/rules.json b/rules.json
deleted file mode 100644
index b870d4cbc2b8e5708213fd4e4f3d1c5501d4467d..0000000000000000000000000000000000000000
--- a/rules.json
+++ /dev/null
@@ -1 +0,0 @@
-{"station":[{"name":"FPGA processing error","interval":"10s","rules":[{"expr":"","for":"20s","labels":{"severity":"major"},"annotations":{"__dashboardUid__":"nC8N_kO7k","__panelId__":"9","summary":"One or more FPGAs are unusable."},"grafana_alert":{"id":1,"orgId":1,"title":"FPGA processing error","condition":"B","data":[{"refId":"A","queryType":"","relativeTimeRange":{"from":600,"to":0},"datasourceUid":"timescaledb","model":{"format":"time_series","group":[],"hide":false,"intervalMs":1000,"maxDataPoints":43200,"metricColumn":"none","rawQuery":true,"rawSql":"SELECT\n $__timeGroup(data_time, $__interval),\n x::text,\n device,\n name,\n value\nFROM lofar_array_boolean\nWHERE\n $__timeFilter(data_time) AND\n name = 'fpga_error_r'\nORDER BY 1,2","refId":"A","select":[[{"params":["value_r"],"type":"column"}]],"table":"att_scalar_devdouble","timeColumn":"data_time","timeColumnType":"timestamp","where":[{"name":"$__timeFilter","params":[],"type":"macro"}]}},{"refId":"B","queryType":"","relativeTimeRange":{"from":0,"to":0},"datasourceUid":"-100","model":{"conditions":[{"evaluator":{"params":[0],"type":"gt"},"operator":{"type":"and"},"query":{"params":["A"]},"reducer":{"params":[],"type":"last"},"type":"query"}],"datasource":{"type":"__expr__","uid":"-100"},"expression":"A","hide":false,"intervalMs":1000,"maxDataPoints":43200,"reducer":"last","refId":"B","settings":{"mode":"dropNN"},"type":"reduce"}}],"updated":"2022-04-04T18:01:53Z","intervalSeconds":10,"version":3,"uid":"kujybCynk","namespace_uid":"R_jsbCynz","namespace_id":6,"rule_group":"FPGA processing error","no_data_state":"NoData","exec_err_state":"Alerting"}}]}]}
\ No newline at end of file
diff --git a/run-wrapper.sh b/run-wrapper.sh
index 3eeb0b0dd65fc806bb22280a063aa031f6c48231..65bc03f5f2b057644a36230018915c352b3df8d7 100755
--- a/run-wrapper.sh
+++ b/run-wrapper.sh
@@ -1,4 +1,6 @@
#! /bin/bash
-/opt/grafana-import/import-rules.sh &
+echo "Starting grafana, with provisioned alarms"
+#/opt/grafana-import/import-rules.sh & # disabled due to incompatibility with alert provisioning
+/opt/grafana-import/populate-tokens.sh
/run.sh
diff --git a/station-dashboards.yaml b/station-dashboards.yaml
new file mode 100644
index 0000000000000000000000000000000000000000..a79059a2b1a46f786a961724ff1d09f5d95c8843
--- /dev/null
+++ b/station-dashboards.yaml
@@ -0,0 +1,45 @@
+apiVersion: 1
+
+providers:
+ # <string> an unique provider name. Required
+ - name: 'StationControl (panels)'
+ # <int> Org id. Default to 1
+ orgId: 1
+ # <string> name of the dashboard folder.
+ folder: ''
+ # <string> folder UID. will be automatically generated if not specified
+ folderUid: ''
+ # <string> provider type. Default to 'file'
+ type: file
+ # <bool> disable dashboard deletion
+ disableDeletion: true
+ # <int> how often Grafana will scan for changed dashboards
+ updateIntervalSeconds: 60
+ # <bool> allow updating provisioned dashboards from the UI
+ allowUiUpdates: true
+ options:
+ # <string, required> path to dashboard files on disk. Required when using the 'file' type
+ path: /var/lib/grafana/panels
+ # <bool> use folder names from filesystem to create folders in Grafana
+ foldersFromFilesStructure: true
+ # <string> an unique provider name. Required
+ - name: 'StationControl (dashboards)'
+ # <int> Org id. Default to 1
+ orgId: 1
+ # <string> name of the dashboard folder.
+ folder: ''
+ # <string> folder UID. will be automatically generated if not specified
+ folderUid: ''
+ # <string> provider type. Default to 'file'
+ type: file
+ # <bool> disable dashboard deletion
+ disableDeletion: true
+ # <int> how often Grafana will scan for changed dashboards
+ updateIntervalSeconds: 60
+ # <bool> allow updating provisioned dashboards from the UI
+ allowUiUpdates: true
+ options:
+ # <string, required> path to dashboard files on disk. Required when using the 'file' type
+ path: /var/lib/grafana/dashboards/station
+ # <bool> use folder names from filesystem to create folders in Grafana
+ foldersFromFilesStructure: true