postgres: Using volumes for persistent data:'initdb: directory "/var/lib/postgresql/data" exists but is not empty'

I managed to run Postgres (after loosing some data and a lot of experimentation) like so (left out some stuff as well):

apiVersion: apps/v1
kind: Deployment
spec:
  template:
    spec:
      containers:
      - name: postgres
        image: postgres:9.5
        env:
        - name: POSTGRES_USER
          value: postgres
        - name: POSTGRES_PASSWORD
          value: "xxxxxxxx"
        - name: POSTGRES_DB
          value: "xxxxxxxx"
        - name: PGDATA
          value: /var/lib/postgresql/data/pgdata
        ports:
        - containerPort: 5432
        volumeMounts:
        - name: postgres-data
          mountPath: "/var/lib/postgresql/data"
      volumes:
      - name: postgres-data
        persistentVolumeClaim:
          claimName: postgres-data

The volume needs to be empty or a valid already initialized postgres database with the file PG_VERSION in there so the init can be skipped.

Did you use a docker volume plugin to create the volume? If there are any files or folders in there like lost+found it will probably fail to initialize. If there are files that you want to keep in the volume (or have no control over) you could adjust the PGDATA environment variable to point to a sub-directory in there like -e PGDATA=/var/lib/postgresql/data/db-files/.

so I managed it by adding a subPath to my Deployment yaml

not 100% sure this is correct, I’m new to this, found on stackoverflow

subPath works

volumeMounts:
  - mountPath: /var/lib/postgresql/data
    name: postgres-storage
    subPath: pgdata # important

also check https://github.com/docker-library/postgres/issues/263#issuecomment-431637087

I actually found what the problem was, and it is indeed not related to the docker env variables. For future references, if the volume is mounted in the POD/Deloypment, don’t export the PGDATA env variable because it will tell initdb to make a new one and to remove the old data.

Running this command gives me the same error: docker run --name some-postgres -v E:\tmp:/var/lib/postgresql/data -d postgres:9.6.15

initdb: directory “/var/lib/postgresql/data” exists but is not empty If you want to create a new database system, either remove or empty the directory “/var/lib/postgresql/data” or run initdb with an argument other than “/var/lib/postgresql/data”.

The folder tmp was not empty and postgres needs an empty folder at start.

Now i get

waiting for server to start…FATAL: data directory “/var/lib/postgresql/data” has wrong ownership

Windows here.

Hi e-nouri, thanks for posting back the fix!. I have the same issue but not the experience to know where to stop the PGDATA env var being exported. How do I do this please as I cannot see it in the pod yml? Is this done in the postgres image (9.6.6) and can be over-ridden by the k8s yml?

Please reopen this, as of now I am using postgres:9.6.6 in Kubernetess 1.8 on GCE, when I updated the cluster I lost all of my data because this does not work.

The wanted behaviour: I mount the volume (persistant claim) to the docker and the database files are stored in the mounted claim, each time I mount that volume, I always find my database there.

The acctual behavior of the postgres:9.6.6 docker:

• Each time the POD boot up, it sets a new database on the mounted claim ! I also exported PGDATA in the deployment file, but it creates a new database each time!

I also ssh’ed into the POD and tested that the PG_VERSION is there and not empty.

The documentation on the docker/_/postgres is confusing and not clear at all.