Using YugabyteDB with Hasura to scale your GraphQL API globally

Software Architect & consultant

Web / Mobile / VR / AR / IoT / AI

GDE, author, engineer

CTO & Co-founder

creating a robust, performant, and feature-rich online conferencing experience

Microservices architecture 

3factor.app Architecture

GraphQL

A query language for your API

What is Hasura

open source and free engine that gives you auto-generated real-time GraphQL API on top of new or existing Postgres database

Features

  • Can be deployed to any cloud or run locally
  • Compatible with all Authentication solutions
  • Can run on top of new or existing Postgres database
  • Supports Postgres addons (PostGIS, TimescaleDB)
  • Auto-generates GraphQL api
  • GraphQL queries are compiled to performant SQL statements using native Postgres features

Features

  • Comes with hasura-cli which has awesome tools like migrations and more
  • Can work with custom SQL statements
  • Has configurable access controls for data
  • Can be connected to your own GraphQL server (does schema stitching)
  • Has eventing system which enables to trigger serverless functions

Let's see it in action

we will run Hasura locally

What about scaling

What about some numbers?

Single instance configuration No. of active live queries CPU load average
1xCPU, 2GB RAM 5000 60%
2xCPU, 4GB RAM 10000 73%
4xCPU, 8GB RAM 20000 90%

Postgres is under about 28% load with peak number of connections being around 850

What is YugabyteDB

open-source, high-performance distributed SQL database for powering global, internet-scale applications.

What is YugabyteDB

Built using combination of

  • high performance document store
  • per-shard distributed consensus replication
  • multi-shard ACID transactions (inspired by Google Spanner)

cloud-native database!

  • public/private clouds
  • Kubernetes environments

What is YugabyteDB

What it brings:

  • PostgreSQL-compatible SQL API
  • low-latency read performance
  • globally distributed write scalability

I will use local Mac install

 
./bin/yb-ctl create --rf 3 --tserver_flags "ysql_suppress_unsupported_error=true"

Create Yugabyte cluster

Cluster Admin UI (localhost:7000)

Run Hasura on top of YugabyteDB

docker run -d -p 8080:8080 -e \
HASURA_GRAPHQL_DATABASE_URL=postgres://postgres:@host.docker.internal:5433/yugabyte \
-e HASURA_GRAPHQL_ENABLE_CONSOLE=true hasura/graphql-engine:latest

Note that port number is 5433

Now let's access Hasura console and see it in action

Run Hasura on top of YugabyteDB

docker run -d -p 8080:8080 -e \
HASURA_GRAPHQL_DATABASE_URL=postgres://postgres:@host.docker.internal:5433/yugabyte \
-e HASURA_GRAPHQL_ENABLE_CONSOLE=true hasura/graphql-engine:latest

Note that port number is 5433

Now let's access Hasura console and see it in action

Let's simulate downtime

 
./bin/yb-ctl stop_node 2

subscriptions will keep running and everything will work as expected

./bin/yb-ctl stop_node 3

2 failed nodes with Replication factor 3

 

Replication factor is 3 so one infrastructure failure is fine

failures that it can tolerate are:

  • disk
  • network
  • availability zone
  • cloud region/datacenter
  • entire cloud

Why one failure is fine, but two are not?

Takeaways

Hasura auto-generate GraphQL API for you

YugabyteDB takes care of high availability of your DB

YugabyteDB global distribution brings the data close to users for multi-region and multi-cloud deployments

Hasura and YugabyteDB are open-source

Start using it today!

Thank You

  @VladimirNovick

consulting/remote workshops/mentoring

Event Loop

Using YugabyteDB with Hasura to scale your GraphQL API globally

By Vladimir Novick

Using YugabyteDB with Hasura to scale your GraphQL API globally

Creating and scaling your GraphQL API can be hard. In this talk we will see not only how we can easily create GraphQL API with Hasura engine, but how we can reduce the common bottleneck - database by bringing high performance distributed SQL database - YugabyteDB. We will see how our GraphQL api will continue to work even in case of catastrophic failures of database nodes, discuss load balancing, how to deploy and manage it and more

  • 1,480