Autoscaling¶

Autoscaling automatically adjusts PgDog settings when the number of processes (e.g. pods in Kubernetes) in the same deployment changes. This simplifies configuration: you no longer need to perform manual calculations to resize connection pools.

Configuration¶

Autoscaling is disabled by default. To enable it, add the following settings to the control plane Helm chart:

values.yaml

control:
  config:
    autoscaling:
      pool_size: true

How it works¶

When a PgDog process connects to the control plane, the control plane provides it with the total number of processes that are part of the same deployment. PgDog then automatically adjusts its configuration by dividing all pool-related configuration values by that number, for example:

Configuration2 processes4 processes8 processes

pgdog.toml

[general]
default_pool_size = 200

[[databases]]
name = "prod"
host = "10.0.0.1"
min_pool_size = 50

users.toml

[[users]]
name = "postgres"
database = "prod"
pool_size = 100

pgdog.toml

[general]
default_pool_size = 100

[[databases]]
name = "prod"
host = "10.0.0.1"
min_pool_size = 25

users.toml

[[users]]
name = "postgres"
database = "prod"
pool_size = 50

pgdog.toml

[general]
default_pool_size = 50

[[databases]]
name = "prod"
host = "10.0.0.1"
min_pool_size = 12

users.toml

[[users]]
name = "postgres"
database = "prod"
pool_size = 25

pgdog.toml

[general]
default_pool_size = 25

[[databases]]
name = "prod"
host = "10.0.0.1"
min_pool_size = 6

users.toml

[[users]]
name = "postgres"
database = "prod"
pool_size = 12

Supported settings¶

The following configuration options are supported for pool size autoscaling:

Section	Configuration
`[general]`	`default_pool_size` (alias: `max_pool_size`)
`[general]`	`min_pool_size`
`[[databases]]`	`pool_size`
`[[databases]]`	`min_pool_size`
`[[users]]`	`pool_size`
`[[users]]`	`min_pool_size`

Orchestrator integration¶

Autoscaling actions are performed entirely using the internal PgDog <-> control plane protocol and, therefore, will work with all orchestrators, including Kubernetes, ECS, and manual deployments.

Kubernetes¶

If using autoscaling and deploying PgDog with our Helm chart, make sure to set the pool-related settings to reflect the total number of connections. For example, if deploying 3 replicas and the total pool size across the 3 pods is 600 connections, set it accordingly in values.yaml:

values.yaml

defaultPoolSize: 600
replicas: 3

users:
  - name: "postgres"
    database: "prod"
    minPoolSize: 300 # 50% of the pool

When PgDog pods are started, they will connect to the control plane and automatically adjust the pool settings to reflect the total number of pods in the deployment:

Configuration	Value	Autoscaled value
`defaultPoolSize`	`600`	`200`
`minPoolSize`	`300`	`100`

Automatic adjustment

The configuration adjustment happens inside the PgDog process. The control plane does not mutate the ConfigMap resource, so GitOps tools like ArgoCD will not detect this drift and continue to operate normally.