Uninstalling Midaz deployment

Fully remove a Midaz Helm deployment from Kubernetes — Helm-managed resources, persistent volumes, secrets, and the namespace itself.

This guide walks you through fully removing a Midaz deployment from Kubernetes. helm uninstall removes the Helm-managed resources — and, importantly, it also deletes the MongoDB PersistentVolumeClaim along with them. Only the PostgreSQL volumes survive, because they are created from a StatefulSet volumeClaimTemplate. This page covers what goes with the release, what stays behind, and how to clean the rest up.

Uninstalling Midaz permanently deletes application data. helm uninstall alone is enough to destroy the MongoDB volume — back up before running it, not after. This operation is irreversible.

What uninstall deletes and what survives

---

Resource	On helm uninstall
Deployments, Services, ConfigMaps, Secrets, ServiceAccounts, RBAC, HPA, PDB, NetworkPolicy	Deleted
Bootstrap Jobs (<release>-bootstrap-postgres, -bootstrap-mongodb, -bootstrap-rabbitmq)	Deleted (they also self-delete 300s after finishing)
PVC <release>-mongodb	Deleted — MongoDB data is lost
PVCs data-<release>-postgresql-primary-0, data-<release>-postgresql-replication-0	Survive (StatefulSet volumeClaimTemplates are not garbage-collected)
RabbitMQ storage	Nothing to delete — RabbitMQ runs on emptyDir, so its data is already gone when the pod stops
Secrets you created yourself (kubectl create secret)	Survive
The namespace	Survives

No chart resource carries a helm.sh/resource-policy: keep annotation. Do not rely on Helm to protect any volume.

Prerequisites

---

Before uninstalling, back up your databases — helm uninstall destroys the MongoDB volume:

kubectl exec -n midaz midaz-mongodb-0 -- mongodump --archive=/tmp/dump.gz --gzip
kubectl cp midaz/midaz-mongodb-0:/tmp/dump.gz ./mongo-backup.gz

The PostgreSQL volumes survive helm uninstall, but the PVC cleanup below destroys them. Dump PostgreSQL too if you intend to run that step:

kubectl exec -n midaz midaz-postgresql-primary-0 -- \
  pg_dumpall -U postgres > ./postgres-backup.sql

Then back up your current Helm values so you can reinstall with the same configuration if needed:

helm get values midaz -n midaz > midaz-values-backup.yaml

Also back up any plugin releases you run separately (plugins install into their own namespace, midaz-plugins):

helm get values plugin-fees -n midaz-plugins > plugin-fees-values-backup.yaml
helm get values plugin-br-pix-direct-jd -n midaz-plugins > plugin-pix-values-backup.yaml

CRM is part of the Midaz chart, not a separate release — there is no plugin-crm chart to back up or uninstall.

Verify all releases that will be affected:

helm list -n midaz
helm list -n midaz-plugins

---

Uninstalling the Helm release

---

Run the following command to uninstall the Midaz Helm release:

helm uninstall midaz -n midaz

This removes all Kubernetes resources created by the Helm chart — Deployments, StatefulSets, Services, Ingresses, ConfigMaps, chart-managed Secrets, ServiceAccounts, RBAC resources, bootstrap Jobs, and the midaz-mongodb PVC. It does not remove the PostgreSQL StatefulSet PVCs, Secrets you created outside Helm, or the namespace.

Verify that all Helm-managed pods have been removed:

kubectl get pods -n midaz

---

Cleaning up persistent resources

---

PersistentVolumeClaims

After uninstall, the PostgreSQL claims are what remains. List them to confirm:

kubectl get pvc -n midaz

You should see data-midaz-postgresql-primary-0 and data-midaz-postgresql-replication-0. The midaz-mongodb claim is already gone — Helm deleted it with the release.

Delete all remaining PVCs in the namespace (this destroys the PostgreSQL data):

kubectl delete pvc --all -n midaz

Or delete specific PVCs by name:

kubectl delete pvc <pvc-name> -n midaz

Deleting these PVCs permanently destroys the PostgreSQL volume data — the ledger’s system of record. Make sure database backups are in place before running this command.

Secrets

Secrets created outside the Helm release lifecycle (e.g., kubectl create secret) are not removed by helm uninstall. List all secrets in the namespace and identify any that are no longer needed:

kubectl get secrets -n midaz

Delete individual orphaned secrets:

kubectl delete secret <secret-name> -n midaz

Or delete all secrets in the namespace:

kubectl delete secrets --all -n midaz

ConfigMaps

ConfigMaps created manually or by bootstrap jobs may also remain. List them:

kubectl get configmaps -n midaz

Delete orphaned ConfigMaps:

kubectl delete configmap <configmap-name> -n midaz

---

Namespace cleanup

---

Once all resources inside the namespace have been removed, delete the namespace itself:

kubectl delete namespace midaz

Deleting the namespace will forcefully remove any remaining resources inside it. If a resource is stuck in Terminating state, you may need to remove its finalizers manually.

Verify the namespace is gone:

kubectl get namespace midaz

---

Complete cleanup

---

For staging, evaluation, or CI environments where a full teardown is safe, the following script automates the entire process:

Data loss is permanent. Run this only in environments where you have confirmed backups or where data loss is acceptable (staging, evaluation, CI). Do not run this in production without a full backup and team sign-off.

#!/bin/bash
set -e

NAMESPACE=midaz
RELEASE=midaz

echo "==> Uninstalling Helm release: $RELEASE"
helm uninstall "$RELEASE" -n "$NAMESPACE" || true

echo "==> Deleting all PersistentVolumeClaims"
kubectl delete pvc --all -n "$NAMESPACE" || true

echo "==> Deleting all Secrets"
kubectl delete secrets --all -n "$NAMESPACE" || true

echo "==> Deleting all ConfigMaps"
kubectl delete configmaps --all -n "$NAMESPACE" || true

echo "==> Deleting namespace: $NAMESPACE"
kubectl delete namespace "$NAMESPACE" || true

echo "==> Done. Midaz has been fully removed."

Save this as midaz-cleanup.sh, make it executable (chmod +x midaz-cleanup.sh), and run it with ./midaz-cleanup.sh.

---

Production considerations

---

In production, a full uninstall requires careful coordination. Follow these steps before running any cleanup commands:

1. Back up all databases before helm uninstall. Export a full snapshot of PostgreSQL and MongoDB — the MongoDB volume is destroyed by the uninstall itself, not by the cleanup steps.
2. Export critical data. If any data needs to be migrated or preserved, export it before uninstalling.
3. Coordinate with your team. Notify all stakeholders of planned downtime and confirm the maintenance window.
4. Uninstall plugins first. Remove plugin releases (Fees, Pix) before uninstalling the core Midaz release. CRM needs no separate uninstall — it ships inside the Midaz chart.
5. Verify no traffic. Confirm that no active traffic is hitting the services before proceeding.

Uninstall plugins before the core release:

# Skips a release that is genuinely absent, and aborts on anything else
# (auth, API, timeout) instead of silently leaving it installed.
# On Helm 3.13+ this is just: helm uninstall <release> -n midaz-plugins --ignore-not-found
for release in plugin-fees plugin-br-pix-direct-jd; do
  if err=$(helm status "$release" -n midaz-plugins 2>&1 >/dev/null); then
    helm uninstall "$release" -n midaz-plugins
  elif printf '%s' "$err" | grep -q 'release: not found'; then
    echo "Skipping $release — not installed."
  else
    echo "Aborting: helm status $release failed: $err" >&2
    exit 1
  fi
done
helm uninstall midaz -n midaz

Then proceed with the persistent resource cleanup steps described above.

---

Related resources

• Deploy Midaz using Helm — Installation guide if you need to reinstall
• Upgrading Midaz and plugins via Helm — Upgrade and rollback procedures
• Troubleshooting — Diagnose issues before deciding to uninstall
• Version compatibility — Version mapping reference