K8s Timezone Config

Kubernetes Pod Timezone Configuration

Standard timezone for Hypera infrastructure: America/Sao_Paulo

Problem

Kubernetes pods run in UTC by default. Logs and application timestamps show +0000 offset instead of local Brazil time (-0300).

Solution

Add the TZ environment variable to container specifications.

Implementation Patterns

1. Helm Values (extraEnv pattern)

For Helm charts that support extraEnv:

# In values.yaml
extraEnv:
  - name: TZ
    value: America/Sao_Paulo

2. Multiple Containers

When a deployment has multiple containers (e.g., API server + frontend), add TZ to ALL containers:

apiServer:
  extraEnv:
    - name: TZ
      value: America/Sao_Paulo

frontend:
  extraEnv:
    - name: TZ
      value: America/Sao_Paulo

3. Raw Kubernetes Deployment

apiVersion: apps/v1
kind: Deployment
spec:
  template:
    spec:
      containers:
        - name: app
          env:
            - name: TZ
              value: America/Sao_Paulo

4. StatefulSet

apiVersion: apps/v1
kind: StatefulSet
spec:
  template:
    spec:
      containers:
        - name: app
          env:
            - name: TZ
              value: America/Sao_Paulo

Verification

After deployment, verify timezone is set correctly:

# Check pod logs for timestamp offset
# Before: 2026-01-13T11:37:06 +0000
# After:  2026-01-13T08:37:06 -0300

# Or exec into pod
kubectl exec -it <pod-name> -n <namespace> -- date
# Should show: Mon Jan 13 08:37:06 -03 2026

Common Applications Requiring Timezone

Important Notes

1. Restart required: Pods must restart for TZ changes to take effect 2. All containers: Set TZ on ALL containers in a pod, including sidecars 3. Init containers: Also set TZ on init containers if they log timestamps 4. Cron jobs: Kubernetes CronJob schedules are always UTC - TZ only affects container-level time

Hypera GitOps Workflow

1. Edit values.yaml in argo-cd-helm-values/kube-addons/<service>/<cluster>/values.yaml 2. Add TZ environment variable to all containers 3. Commit and push (ArgoCD auto-syncs) 4. Verify pods restart with new timezone

Reference

• IANA Timezone Database: America/Sao_Paulo = UTC-3 (no DST since 2019)
• Linux TZ variable: Uses /usr/share/zoneinfo/America/Sao_Paulo

---

Gotchas

• `TZ` env var only affects app-level timestamps, not kubectl logs timestamps (those come from the container runtime, which stays UTC). The kubelet timestamps stay +0000 even after the pod's internal clock is São Paulo.
• CronJob `schedule:` is always UTC regardless of pod TZ. A 0 8 * * * schedule fires at 05:00 BRT, not 08:00 BRT. Use spec.timeZone: America/Sao_Paulo on the CronJob itself (K8s 1.27+) or compute the UTC offset manually.
• Distroless and scratch images have no `/usr/share/zoneinfo` — TZ=America/Sao_Paulo silently falls back to UTC. Either use a base image that ships tzdata or mount it via configMap/volume.
• Init containers don't inherit `TZ` from the main container in older charts — set it on the init container spec separately or their migration logs stay UTC.
• Java apps need `-Duser.timezone=America/Sao_Paulo` in JAVA_OPTS in addition to TZ — the JVM reads its own property, not the env var, on some distributions.
• PostgreSQL `TZ` env var sets the OS clock, not the database timezone setting. Server timestamps via now() still use the value in postgresql.conf — set both.