Configuration

Settings you may want to change

absolutely set these before start the instance for the first time:

NEODB_SECRET_KEY - 50 characters of random string, no white space
NEODB_SITE_NAME - the name of your site
NEODB_SITE_DOMAIN - the domain name of your site

NEODB_SECRET_KEY and NEODB_SITE_DOMAIN must not be changed later.

if you are doing debug or development:

NEODB_DEBUG - True will turn on debug for both neodb and takahe, turn off relay, and reveal self as debug mode in nodeinfo (so peers won't try to run fedi search on you)
NEODB_IMAGE - the docker image to use, neodb/neodb:edge for the main branch

Settings for Federating

DISABLE_DEFAULT_RELAY is set to False by default, the server will send and receive public posts from relay.neodb.net. relay.neodb.net is open sourced and operated by NeoDB developers, it works like most ActivityPub relays except it only relays between NeoDB instances, it helps public information like catalogs and trends flow between NeoDB instances. You may set it to True if you don't want to relay public posts with other NeoDB instances.

Settings for Scrapers

TBA

Other maintenance tasks

Add alias to your shell for easier access

alias neodb-manage='docker-compose --profile production run shell neodb-manage'

Enable Developer Console

neodb-manage createapplication --client-id NEODB_DEVELOPER_CONSOLE --skip-authorization --name 'NeoDB Developer Console' --redirect-uris 'https://example.org/lol'  confidential authorization-code

Multiple instances on one server

It's possible to run multiple clusters in one host server, as long as NEODB_SITE_DOMAIN, NEODB_PORT and NEODB_DATA are different.

Scaling up

For high-traffic instance, spin up these configurations to a higher number, as long as the host server can handle them:

NEODB_WEB_WORKER_NUM
NEODB_API_WORKER_NUM
NEODB_RQ_WORKER_NUM
TAKAHE_WEB_WORKER_NUM
TAKAHE_STATOR_CONCURRENCY
TAKAHE_STATOR_CONCURRENCY_PER_MODEL

Further scaling up with multiple nodes (e.g. via Kubernetes) is beyond the scope of this document, but consider run db/redis/typesense separately, and then duplicate web/worker/stator containers as long as connections and mounts are properly configured; migration only runs once when start or upgrade, it should be kept that way.

2.4 KiB Raw Blame History