nais / api Goto Github PK

Earlier today, for a team which used more than requested of cpu/memory, the following query would result in an error:

query Overview($team: Slug!) {
  currentResourceUtilizationForTeam(team: $team) {
    cpu {
      utilization
      estimatedAnnualOverageCost
    }
    memory {
      utilization
      estimatedAnnualOverageCost
    }
  }
  resourceUtilizationTrendForTeam(team: $team) {
    cpuUtilizationTrend
    memoryUtilizationTrend
  }
}

Both currentResourceUtilizationForTeam and resourceUtilizationTrendForTeam would fail because of the following query would return no rows:

SELECT
    SUM(usage)::double precision AS usage,
    SUM(request)::double precision AS request,
    timestamp
FROM
    resource_utilization_metrics
WHERE
    team_slug = @team_slug
    AND resource_type = @resource_type
    AND timestamp = @timestamp
    AND request > usage
GROUP BY
    timestamp;

The problem is likely AND request > usage which was added to solve nais/console-backend#40

I made a workaround in 0a5cd99 which returns empty data when there's no rows.

ref:

https://nav-it.slack.com/archives/C5KUST8N6/p1707299594408229

When deploying an application, when new instances are starting up they are reported as "FAILING". This results in the app itself is reported as failing, which is not correct. It's just starting up.

This should be corrected so that users don't think something is wrong when it's not

Alle medlemmer av et team, ikke bare owners bør kunne endre slack alert-kanaler.

• Remove reconciler resources from graph
• Reconciler resources: Value fra text til blob
  Trenger vi dette?
• Eksponere måter å oppdatere environment og team på
  Team er done
• Flytte slack alerts channels til environments
• GRPC for reconciler_errors
• Teste OAuth flyten

Har ikke sjekket dagens kode, men mulig det fortsatt er slik..

ref: https://nav-it.slack.com/archives/C050DP53VPH/p1707295683898859

• Bytte epost på nais-console@tenantdomain til nais-admin.
• Etter at vi setter opp, så vil en ny gruppe for admins som heter nais-admins. Den gamle console-admins kan da fjernes

Workloads not configuring any resource requirements in manifest is not showing any cost.
Utilization is still reported.

Example: https://console.nav.cloud.nais.io/team/tbd/prod-gcp/app/spekemat-slakter/manifest

apiVersion: nais.io/v1alpha1
kind: Application
metadata:
  labels:
    team: tbd
  name: spekemat-slakter
  namespace: tbd
spec:
  accessPolicy:
    outbound:
      rules:
      - application: spekemat
  azure:
    application:
      enabled: true
      tenant: nav.no
  env:
  - name: KAFKA_RAPID_TOPIC
    value: tbd.rapid.v1
  - name: KAFKA_CONSUMER_GROUP_ID
    value: tbd-spekemat-v1
  - name: KAFKA_RESET_POLICY
    value: latest
  image: europe-north1-docker.pkg.dev/nais-management-233d/tbd/helse-spekemat-slakter:2024.02.16-22.10-c1252ac
  kafka:
    pool: nav-prod
  liveness:
    path: /isalive
  preStopHook:
    http:
      path: /stop
  prometheus:
    path: /metrics
  readiness:
    path: /isready
  replicas:
    max: 24
    min: 2
    scalingStrategy:
      kafka:
        consumerGroup: tbd-spekemat-v1
        threshold: 200
        topic: tbd.rapid.v1
  secureLogs:
    enabled: true
  strategy:
    type: RollingUpdate

When not specifying resource-specs, each pod will default to this:

    resources:
      limits:
        memory: 512Mi
      requests:
        cpu: 200m
        memory: 256Mi

This should at least be visible cost-wise, but also informed clearly that these values are defaulted to when not configured.