POSTGRES.md

PostgreSQL requirements for Daml Enterprise Helm Charts deployment

Any PostgreSQL server that meets these criteria is compatible.

Example SaaS solutions:

• Amazon RDS for PostgreSQL or Aurora PostgreSQL
• Azure Database for PostgreSQL
• Google Cloud SQL or Cloud Spanner

Table of contents

• Supported versions
• Minimum server size
• Availability
• Concurrent connections
• SSL/TLS
• Debugging

---

Supported versions

PostgreSQL server versions 11 to 14 are supported by Canton and Daml services.

Minimum server size

If you install each Helm chart once (Domain, Participant, HTTP JSON API service & Trigger service) using a shared PostgreSQL server for all databases, you should start with 4 vCPUs/cores and 16Go RAM. Check the storage to make sure you have enough space, IOPS and throughput available.

Availability

Your PostgresQL server should be highly available. Each Canton/Daml component always needs to be connected to its own database.

Concurrent connections

You should verify your PostgreSQL server maximum allowed connections, espically if you share the same instance for all components, to allow all of them to connect without issues.

The open-source PostgreSQL server default is 100, a good start is to set max_connections to at least 200. Remember that you also need to have enough room for system (WAL, autovacuum, replication, etc.) and admin connections.

On most cloud providers' SaaS solution, the default server max connections is scaled according to your instance size and its available memory.

Client side settings are also available in Helm values, you can find them in the Parameters section of each chart readme. For Canton nodes you can set storage.maxConnections. For Daml services you can set storage.poolSize and storage.minIdle.

An HA deployment of all the Helm charts in this repository with the default values requires around 120 connections:

• Single Domain with two Sequencers
• Single Participant
• Single HTTP JSON API service
• Single Trigger service

TLS

We strongly recommend to enable and enforce TLS for connections to your PostgreSQL server (Helm charts default). Please refer to the relevant PostgreSQL server and client documentation.

On most cloud providers' SaaS solution, server SSL/TLS is enabled by default and certificates are signed by a public CA. In your storage values, sslRootCert, sslCert and sslKey can be left with an empty string value "". The default JVM trust store with common public CA certificates will be used.

Scaling

In case your PostgreSQL server load gets high you can either or both:

• Scale vertically using a bigger server with more CPU/RAM and faster storage
• Scale horizontally and host each database on a different PostgreSQL server

To achieve high performance under heavy workload, make sure your server has enough CPU/RAM and its storage has enough space, IOPS and throughput available.

On most cloud providers' SaaS solution, the default storage max IOPS and max throughput are scaled according to its size, there might also be different tiers/types.

---

Debugging

A few useful PostgreSQL queries below in case of issues with your server and/or databases.

Check locks

SELECT * FROM pg_locks;

Display database IDs and connection information

SELECT oid AS db_id,
       datname AS db_name,
       datallowconn AS db_allow_connect,
       datconnlimit AS db_connection_limit
FROM pg_database
ORDER BY oid;

All requests by specific user

Using psql can list roles/users with \du+ then:

SELECT * FROM pg_stat_activity WHERE usename='<role_name>';

Display server max connections setting

show max_connections;

Total connections to server

SELECT sum(numbackends) FROM pg_stat_database;

Total connections to server per role/user

SELECT count(*) as connections, usename FROM pg_stat_activity GROUP BY usename ORDER BY connections DESC;