Skip to content

Latest commit

 

History

History
238 lines (184 loc) · 7.1 KB

developer.md

File metadata and controls

238 lines (184 loc) · 7.1 KB
Raw

Overview

BCDHub is a set of microservices written in Golang:

  • indexer
    Loads and decodes operations related to smart contracts and also keeps track of the block chain and handles protocol updates.
  • metrics
    Receives new contract/operation events from the indexer and calculates various metrics that are used for ranking, linking, and labelling contracts and operations.
  • API
    Exposes RESTful JSON API for accessing indexed data (with on-the-fly decoding). Also provides a set of methods for authentication and managing user profiles.

Those microservices are sharing access to databases and communicating via message queue:

  • ElasticSearch cluster (single node) for storing all indexed data including blocks, protocols, contracts, operations, Big_map diffs, and others.
  • PostgreSQL database for storing compilations and user data.
  • RabbitMQ for communications between API, indexer and metrics.

Third-party services

BCDHub also depends on several API endpoints exposed by TzKT although they are optional:

  • List of blocks containing smart contract operations, used for boosting the indexing process (allows to skip blocks with no contract calls)
  • Mempool operations
  • Contract aliases and other metadata

Those services are obviously make sense for public networks only and not used for sandbox or other private environments.

Versioning

BCD uses X.Y.Z version format where:

  • X changes every 3-5 months along with a big release with a significant addition of functionality
  • Y increasing signals about a possibly non-compatible update that requires reindexing (or restoring from snaphot) or syncing with frontend
  • Z bumped for every stable release candidate or hotfix

Syncing with frontend

BCD web interface developed at https://github.com/baking-bad/bcd uses the same version scheme.
X.Y.* versions of backend and frontent MUST BE compatible which means that for every change in API responses Y has to be increased.

Publishing releases

Is essentially tagging commits:

make release  # forced tag update

For stable release:

git tag X.Y.Z
git push --tags

Docker images

Although you can install and run each part of BCD Hub independently, as system services for instance, the simplest approach is to use dockerized versions orchestrated by docker-compose.

BCDHub docker images are being built on dockerhub. Tags for stable releases have format X.Y.

Linking with Git tags

Docker tags are essentially produced from Git tags using the following rules:

  • X.Y.* → X.Y

Building images

make images  # latest
make stable-images  # requires STABLE_TAG variable in the .env file

Deploy

Requirements

Make sure you have installed:

  • docker
  • docker-compose

You will also need several ports to be not busy:

  • 14000 API service
  • 9200 Elastic
  • 5672 RabbitMQ
  • 5432 PostgreSQL
  • 8000 Frontend GUI

Get ready

  1. Clone this repo
git clone https://github.com/baking-bad/bcdhub.git
cd bcdhub
  1. Create and fill .env file (see Configuration)
your-text-editor .env

Environments

There are several predefined configurations serving different purposes.

Production better-call.dev

  • Stable docker images X.Y
  • /configs/production.yml file is used internally
  • Requires STABLE_TAG environment set
  • Deployed via make stable

Staging you.better-call.dev

  • Latest docker images latest
  • /configs/you.yml file is used internally
  • Deployed via make latest

Development localhost

  • /configs/development.yml file is used
  • You can spawn local instances of databases and message queue or ssh to staging host with port forwarding
  • Run services make {service} (where service is one of api indexer metrics)

Sandbox bbbox

  • /configs/sandbox.yml file is used
  • Start via COMPOSE_PROJECT_NAME=bcd-box docker-compose -f docker-compose.sandbox.yml up -d --build
  • Stop via COMPOSE_PROJECT_NAME=bcd-box docker-compose -f docker-compose.sandbox.yml down

Running

Startup

It takes around 20-30 seconds to initialize all services, API endpoints might return errors until then.
NOTE that if you specified local RPC node that's not running, BCDHub will wait for it indefinitely.

Snapshots

Full indexing process requires about 2 hours, however there are cases when you cannot afford that.
Elastic Search has a built-in incremental snapshotting mechanism which we use together with the AWS S3 plugin.

NOTE: currently we don't provide public snapshots.
You can set up your own S3 repo: https://medium.com/@federicopanini/elasticsearch-backup-snapshot-and-restore-on-aws-s3-f1fc32fbca7f
Alternatively, contact us for granting access

Get ready

  • Make sure you have snapshot settings in your .env file
  • Elastic service should be up and initialized

Make snapshot

1. Initialize credentials

make s3-creds

No further actions required

2. Create local repository (if not exists)

make s3-repo

Follow the instruction: you can choose an arbitrary name for your repo.

3. Create snapshot

make s3-snapshot

Select an existing repository to store your snapshot.

4. Schedule automatic snapshots

make s3-policy

Select an existing repository and configure time intervals using cron expressions: https://www.elastic.co/guide/en/elasticsearch/reference/master/trigger-schedule.html#schedule-cron

Restore snapshot

1. Cleanup (optional)

In some cases it's not possible to apply snapshot on top of existing indices. You'd need to clear the data then.
WARNING: This will delete all data from your elastic instance.

make es-reset

Wait for Elastic to be initialized.

2. Initialize creds and repo

Follow steps 1 and 2 from the make snapshot instruction.

3. Apply snapshot

make s3-restore

Select the latest (by date) snapshot from the list. It's taking a while, don't worry about the seeming freeze.

Version upgrade

This is mostly for production environment, for all others a simple "start from the scratch" would work.

Soft update

E.g. applying hotfixes. No breaking changes in the database schema.

1. Build stable images

Make sure you are on master branch

git pull
make stable-images
make stable

1'. Pull stable images

make stable-pull

2. Deploy

make stable

Data migration

E.g. new field added to one of the elastic models. You'd need to write a migration script to update existing data.

1. Pull migration script

git pull

2. Run migration

make migration

Select your script.

Upgrade from snapshot

In case you need to reindex from scratch you can set up a secondary BCDHub instance, fill the index, make a snapshot, and then apply it to the production instance.

0. Make a snapshot

Typically you'd use staging for that.

1. Stop BCDHub and clear indexed data

make upgrade

Wait for Elastic to be initialized after restart.

2. Restore snapshot

make s3-restore

Select the snapshot you made.

3. Run the rest of the services

make stable