Need guidance for running the complete Supabase stack natively on Ubuntu (without Docker) #47908

Gopendra-cpu · 2026-07-14T07:14:21Z

Gopendra-cpu
Jul 14, 2026

I am trying to run the complete Supabase stack without Docker on an Ubuntu 22.04 server. My goal is to install and manage each service natively using systemd.

Current Setup

My existing production environment is running with Docker:

PostgreSQL (supabase/postgres:15.8.1.085)
GoTrue (supabase/gotrue:v2.186.0)
PostgREST (postgrest/postgrest:v14.8)
Postgres Meta
Kong

On the new server:

Ubuntu 22.04.5 LTS
PostgreSQL 15.18 installed natively
GoTrue successfully built from source and runs correctly
PostgREST binary downloaded
PostgreSQL backup available (pg_dumpall SQL file)
Issue

When restoring the PostgreSQL backup into a native PostgreSQL installation, I receive errors about missing Supabase-specific objects, for example:

schema "net" does not exist
relation "storage.objects" does not exist
function extensions.algorithm_sign(...) does not exist

From inspecting the old Docker container, I found it includes extensions such as:

pg_graphql
pg_net
pgjwt
pgcrypto
supabase_vault
uuid-ossp
pg_stat_statements

It also uses PostgreSQL settings like:

wal_level = logical
shared_preload_libraries including pg_net, pgsodium, pg_cron, supabase_vault, etc.

I also cloned the supabase/postgres repository and checked out the 15.8.1.085 tag. I noticed the project uses Docker, Nix, and Ansible to build PostgreSQL, but I couldn't find installable PostgreSQL packages or documentation for installing the same environment directly on Ubuntu.

Kelbimon · 2026-07-14T08:44:07Z

Kelbimon
Jul 14, 2026

A pg_dumpall restore is not enough to reproduce a Supabase PostgreSQL installation.

supabase/postgres:15.8.1.085 and a normal PostgreSQL 15 installation are not equivalent. The Supabase image includes specific extension binaries, control files, configuration, roles, and initialization SQL. The tagged Dockerfile installs a versioned Nix-built PostgreSQL toolchain and extensions, then copies Supabase configuration and database migrations into the image:

https://github.com/supabase/postgres/blob/15.8.1.085/Dockerfile-15

A PostgreSQL dump contains SQL definitions and data. It does not contain native extension libraries such as pg_net.so, their control files, or preload configuration. If the dump contains something like:

CREATE EXTENSION pg_net;

but pg_net is not installed on the target server, that statement fails. The later errors are often just consequences:

pg_net creates the net schema, so a failed extension installation leads to schema "net" does not exist.
extensions.algorithm_sign(...) comes from pgjwt, so it will be missing if that extension was not installed successfully.
In this Supabase Postgres release, storage.objects is created by the image's initialization SQL. Its creation depends on other Supabase objects, including auth.users and extensions.uuid_generate_v4(). If an earlier dependency fails, later references to storage.objects will also fail.

The storage initialization script for this exact release is here:

https://github.com/supabase/postgres/blob/15.8.1.085/migrations/db/init-scripts/00000000000002-storage-schema.sql

I would first restore again with error stopping enabled:

psql -X --set ON_ERROR_STOP=on -f backup.sql postgres

That identifies the first real failure instead of continuing and producing hundreds of secondary errors.

I would also compare the source and target databases:

SELECT version();

SELECT extname, extversion
FROM pg_extension
ORDER BY extname;

SHOW shared_preload_libraries;

To run PostgreSQL natively, you would need to reproduce the database environment from the tagged Supabase image:

Build and install the same extension versions used by 15.8.1.085.
Install each extension's .so, .control, and SQL files into the directories used by PostgreSQL 15.
Reproduce the required PostgreSQL configuration, including shared_preload_libraries, wal_level, extension settings, and preload ordering.
Recreate the Supabase roles and permissions.
Apply the tagged Supabase initialization SQL in the correct order.
Run the matching service migrations for Auth, Storage, and the other services.
Only then restore the application data.

GoTrue, PostgREST, Postgres Meta, and Kong are also not the entire Supabase stack. Depending on the features in use, the deployment may also require Storage API, Realtime, Supavisor, Functions, imgproxy, Logflare/Vector, and other services.

The official self-hosting documentation currently recommends Docker, and I could not find an official procedure for installing this release as native Ubuntu packages managed by systemd:

https://supabase.com/docs/guides/self-hosting

A native installation is possible, but it means maintaining a custom PostgreSQL distribution yourself. The Nix and Ansible files in supabase/postgres are the closest available build specification; they are not a ready-made Ubuntu package.

For a production migration, I would keep at least PostgreSQL on the matching supabase/postgres:15.8.1.085 image unless there is a strong reason not to. The other services can still be managed independently, but replacing the database image requires reproducing all of its compiled extensions and bootstrap behavior, not just installing PostgreSQL 15 from Ubuntu's packages.

1 reply

Gopendra-cpu Jul 20, 2026
Author

Hi,

First of all, thank you for taking the time to explain everything in such detail. Your previous explanation helped us understand the root cause much better.

From what I understood from your message:

A pg_dumpall backup alone is not sufficient to reproduce a Supabase PostgreSQL installation.
The supabase/postgres:15.8.1.085 image contains much more than a standard PostgreSQL 15 installation. It includes:
- Supabase-specific PostgreSQL extensions
- Extension binaries (.so), control files, and SQL files
- PostgreSQL configuration
- Required roles and permissions
- Bootstrap/initialization SQL
- Service migrations
- Because of this, simply restoring the backup into a native PostgreSQL installation results in many missing dependency errors.

Based on your explanation, we started investigating in that direction.

What we tried

Installed PostgreSQL 15 natively on Ubuntu.
Restored the pg_dumpall backup.
Analysed the restore log to identify the first failures.

The first errors we found were missing extensions such as:

pg_net
pg_graphql
pgjwt
supabase_vault

After those failures, we started getting many secondary errors such as:

Missing schemas (net, storage, vault)
Missing functions
Missing tables like storage.objects
Role/permission errors (supabase_admin, supabase_auth_admin, etc.)
PostgreSQL configuration issues (shared_preload_libraries, wal_level, event triggers, etc.)

So it now seems clear that the restore expects a fully initialized Supabase PostgreSQL environment before importing the backup.

Where we're currently stuck

At this point, we're trying to decide the best execution plan.

One approach we're considering is to build the native Supabase PostgreSQL environment first, then restore the backup only after the environment matches the original Docker image.

We'd really appreciate your opinion on whether this is the right direction.

A few questions

Do you think building the native Supabase PostgreSQL environment is the correct approach, or would you recommend a different strategy?
If you were implementing this migration, what order would you follow?
How would you recreate all the required extensions, roles, schemas, configuration, and bootstrap SQL outside Docker?
Is there any official or community-supported process for installing supabase/postgres:15.8.1.085 natively on Ubuntu?
Are there any hidden dependencies or common pitfalls that we should be aware of before continuing?
If our goal is to run Supabase without Docker, would you recommend building the PostgreSQL environment from the supabase/postgres source repository, or is there a better alternative?
If you were doing this migration yourself, what would be your overall execution plan from start to finish?

If you've worked on a similar migration before, we'd be very grateful if you could share your experience or any best practices.

Thank you again for your time and guidance. Your previous explanation has already helped us understand the problem much more clearly, and any additional suggestions or recommendations would be greatly appreciated.

satiricalguru · 2026-08-28T11:04:36Z

satiricalguru
Aug 28, 2026

The reason your pg_dumpall restore fails with schema "net" does not exist, relation "storage.objects" does not exist, and function algorithm_sign does not exist is that Supabase PostgreSQL depends on several custom C extensions, predefined roles, and system schemas (auth, storage, vault, net, extensions) that do not exist in a stock apt install postgresql-15 installation.

Here is the step-by-step procedure to configure native Ubuntu 22.04 PostgreSQL so that your dump restores cleanly.

Step 1: Install Build Dependencies & PostgreSQL Dev Headers

sudo apt update
sudo apt install -y build-essential git cmake pkg-config postgresql-server-dev-15 \
                    libsodium-dev libcurl4-openssl-dev libssl-dev libpq-dev

Step 2: Build & Install Missing Supabase Extensions

1. `pgjwt` (Provides `algorithm_sign` and JWT signing)

git clone https://github.com/michelp/pgjwt.git
cd pgjwt
sudo make install
cd ..

2. `pg_net` (Provides `schema "net"` for async networking)

git clone https://github.com/supabase/pg_net.git
cd pg_net
sudo make install
cd ..

3. `pgsodium` & `supabase_vault` (Encryption and secrets management)

git clone https://github.com/michelp/pgsodium.git
cd pgsodium
sudo make install
cd ..

git clone https://github.com/supabase/vault.git
cd vault
sudo make install
cd ..

4. `pg_graphql` & `pgvector`

# pgvector
git clone --branch v0.7.4 https://github.com/pgvector/pgvector.git
cd pgvector
sudo make install
cd ..

# pg_cron
sudo apt install -y postgresql-15-cron

Step 3: Configure `postgresql.conf`

Edit /etc/postgresql/15/main/postgresql.conf:

# Add Supabase preload libraries
shared_preload_libraries = 'pg_stat_statements, pg_cron, pg_net, pgsodium, supabase_vault'

# Logical replication for Realtime
wal_level = logical

# Custom schema search paths
search_path = '"$user", public, extensions'

Restart PostgreSQL:

sudo systemctl restart postgresql

Step 4: Bootstrap Core Roles & System Schemas

Before restoring the dump, you must initialize the Supabase standard roles and system schemas. Connect as the postgres superuser:

-- 1. Create standard Supabase roles
CREATE ROLE anon NOLOGIN;
CREATE ROLE authenticated NOLOGIN;
CREATE ROLE service_role NOLOGIN;
CREATE ROLE supabase_admin LOGIN SUPERUSER PASSWORD 'your-secure-password';
CREATE ROLE authenticator LOGIN NOINHERIT PASSWORD 'your-authenticator-password';

GRANT anon TO authenticator;
GRANT authenticated TO authenticator;
GRANT service_role TO authenticator;

-- 2. Create required system schemas
CREATE SCHEMA IF NOT EXISTS extensions;
CREATE SCHEMA IF NOT EXISTS auth;
CREATE SCHEMA IF NOT EXISTS storage;
CREATE SCHEMA IF NOT EXISTS vault;
CREATE SCHEMA IF NOT EXISTS graphql_public;

-- 3. Enable extensions inside the extensions schema
CREATE EXTENSION IF NOT EXISTS "uuid-ossp" SCHEMA extensions;
CREATE EXTENSION IF NOT EXISTS "pgcrypto" SCHEMA extensions;
CREATE EXTENSION IF NOT EXISTS "pgjwt" SCHEMA extensions;
CREATE EXTENSION IF NOT EXISTS "pg_net" SCHEMA extensions;
CREATE EXTENSION IF NOT EXISTS "pg_cron" SCHEMA extensions;
CREATE EXTENSION IF NOT EXISTS "pgsodium" SCHEMA extensions;
CREATE EXTENSION IF NOT EXISTS "supabase_vault" SCHEMA extensions;

Step 5: Restore Your Dump

Now that the schemas, roles, and extension C-libraries exist in your native Postgres instance, restore your backup:

# Restore schemas and data
psql -U postgres -d postgres -f /path/to/your_backup.sql

After the restore succeeds, you can run gotrue and postgrest binaries against this native database via systemd without missing relation/function errors.

0 replies

Navigation Menu

Supabase

Need guidance for running the complete Supabase stack natively on Ubuntu (without Docker) #47908

Uh oh!

{{title}}

Uh oh!

Replies: 2 comments 1 reply

Uh oh!

{{title}}

Uh oh!

Uh oh!

{{title}}

Uh oh!

Uh oh!

{{title}}

Uh oh!

Select a reply

Uh oh!

Supabase

Need guidance for running the complete Supabase stack natively on Ubuntu (without Docker) #47908

Uh oh!

Gopendra-cpu Jul 14, 2026

Replies: 2 comments · 1 reply

Uh oh!

Kelbimon Jul 14, 2026

Uh oh!

Gopendra-cpu Jul 20, 2026 Author

Uh oh!

satiricalguru Aug 28, 2026

Step 1: Install Build Dependencies & PostgreSQL Dev Headers

Step 2: Build & Install Missing Supabase Extensions

1. pgjwt (Provides algorithm_sign and JWT signing)

2. pg_net (Provides schema "net" for async networking)

3. pgsodium & supabase_vault (Encryption and secrets management)

4. pg_graphql & pgvector

Step 3: Configure postgresql.conf

Step 4: Bootstrap Core Roles & System Schemas

Step 5: Restore Your Dump

Gopendra-cpu
Jul 14, 2026

Replies: 2 comments 1 reply

Kelbimon
Jul 14, 2026

Gopendra-cpu Jul 20, 2026
Author

satiricalguru
Aug 28, 2026

1. `pgjwt` (Provides `algorithm_sign` and JWT signing)

2. `pg_net` (Provides `schema "net"` for async networking)

3. `pgsodium` & `supabase_vault` (Encryption and secrets management)

4. `pg_graphql` & `pgvector`

Step 3: Configure `postgresql.conf`