Slide 1
Slide 1 text
Introduction and Insights of the Hasura-based Architecture
Meetup #21 Guest talk and lightning talks
2023.09.13
Slide 2
Slide 2 text
Kazuki Hayaseʢ@KazukiHayaseʣ
● Frontend Engineer at BuySell Technologies
● Currently working on the management system for
listing items on e-commerce websites
● GraphQL / TypeScript / Golang
Slide 3
Slide 3 text
Introduction of the Hasura-based architecture
and pros/cons
Topic
Slide 4
Slide 4 text
Agenda
What is Hasura?
01
Overall Architecture
02
Pros and Cons
03
Conclusion
04
Slide 5
Slide 5 text
01
What is Hasura?
Slide 6
Slide 6 text
● OSS that automatically generates a GraphQL schema from a
database schema
● Once the tables are created, CRUD APIs get generated. So a
GraphQL server can be built quickly
● Authentication, authorization and federation are also available
What is Hasura?
Slide 7
Slide 7 text
What is Hasura?
Table creation Auto-generated GraphQL schema
Slide 8
Slide 8 text
● Can be authenticated and authorized by integrating with many
popular auth services
● Has a role-based authorization control at row and column level
Authentication and Authorization
Slide 9
Slide 9 text
Hasura has the ability to integrate GraphQL and REST endpoints
● Remote Schemas
● Actions
Federation
Slide 10
Slide 10 text
Federation
Remote Schema is a feature that merges your remote GraphQL
schemas and provides you with a single, unified GraphQL API
Slide 11
Slide 11 text
02
Overall Architecture
Slide 12
Slide 12 text
Overall Architecture
Slide 13
Slide 13 text
● Builds a GraphQL server using Golang
○ Merges this GraphQL server into Hasura as Remote Schema
● Using gqlgen for schema-first development
● Complex Queries and Mutations are implemented on this server
Custom GraphQL Server
Slide 14
Slide 14 text
● Adopted Elasticsearch to meet the requirements for complex
searching
● Implemented integration with Elasticsearch in Remote Schema
○ As of today (September 13th, 2023), Hasura does not support Elasticsearch
Elasticsearch
Slide 15
Slide 15 text
● In Remote Schema, it returns only the 'id'
● In Hasura, fetching data from the database based on the 'id'
Elasticsearch
Slide 16
Slide 16 text
03
Pros and Cons
Slide 17
Slide 17 text
Pros
01 The implementation effort has been reduced
02 Hasura takes on the role of the Federation Layer
03 Configurations can be managed through the GUI
Slide 18
Slide 18 text
Pros
01 The implementation effort has been reduced
02 Hasura takes on the role of the Federation Layer
03 Configurations can be managed through the GUI
Slide 19
Slide 19 text
● Simple Queries and Mutations can use auto-generated APIs
● There is no need to implement resolvers for those parts
● Overall, auto-generated APIs are used for 90% of Queries and
10% of Mutations
○ Due to the frequent external integration, the proportion of mutations is low
The implementation effort has been reduced
Slide 20
Slide 20 text
Pros
01 The implementation effort has been reduced
02 Hasura takes on the role of the Federation Layer
03 Configurations can be managed through the GUI
Slide 21
Slide 21 text
● Hasura federates the REST and GraphQL endpoints
● Federation setting is easily managed through the GUI
● Authorization for the federated target can also be managed with
Hasura
Hasura takes on the role of the Federation Layer
Slide 22
Slide 22 text
Hasura takes on the role of the Federation Layer
Federation setting Authorization setting
Slide 23
Slide 23 text
Pros
01 The implementation effort has been reduced
02 Hasura takes on the role of the Federation Layer
03 Configurations can be managed through the GUI
Slide 24
Slide 24 text
● Most settings can be managed through the GUI
● The settings are exported as YAML files
● Database migrations can also be managed through the GUI
Configurations can be done through the GUI
Slide 25
Slide 25 text
Cons
01 Managing configuration files can be difficult
02 Knowledge of the tables is required to write queries
03 Tuning the database is difficult
Slide 26
Slide 26 text
Cons
01 Managing configuration files can be difficult
02 Knowledge of the tables is required to write queries
03 Tuning the database is difficult
Slide 27
Slide 27 text
● Reviewing the YAML configuration files alone is difficult
● In large-scale systems, managing configs through the GUI can
actually be more difficult
Managing configuration files can be difficult
Slide 28
Slide 28 text
● Making rules related to configuration
● Creating a custom YAML file validator based on the rules and
running it in CI
● Additionally, creating a custom tool for modifying YAML files
Solutions
Slide 29
Slide 29 text
Cons
01 Managing configuration files can be difficult
02 Knowledge of the tables is required to write queries
03 Tuning the database is difficult
Slide 30
Slide 30 text
Knowledge of the tables is required to write queries
With Hasura, the step of defining the GraphQL schema is omitted
Usually
Using Hasura
Table definition GraphQL schema
definition
Implementation
Table definition Implementation
Slide 31
Slide 31 text
Knowledge of the tables is required to write queries
Understanding both the GraphQL schema and
the table structure is necessary
Writing Queries becomes dependent on individual skills
Slide 32
Slide 32 text
● Participating in backend development to get the knowledge of
the table structure
○ The team's policy of developing in full-stack is the underlying principle
● Export only the required schema
○ Reducing the cost of schema selection
Solutions
Slide 33
Slide 33 text
Cons
01 Managing configuration files can be difficult
02 Knowledge of the tables is required to write queries
03 Tuning the database is difficult
Slide 34
Slide 34 text
● Hasura builds responses using SQL queries, often resulting in
complex SQL queries
● Understanding the complex SQL queries is nearly impossible
● It is not possible to modify the SQL query unless fixing the
GraphQL Query
Tuning the database is difficult
Slide 35
Slide 35 text
Tuning the database is difficult
GraphQL Query Corresponding SQL query
Slide 36
Slide 36 text
● We can only monitor slow queries and add indexes
● If this doesn't solve the performance issue, the only option is to
implement that Query in the Remote Schema
Solutions
Slide 37
Slide 37 text
04
Conclusion
Slide 38
Slide 38 text
● Adopting Hasura has reduced development efforts and allowed
us to focus on core logic implementation
○ It's highly recommended for Proof of Concept or creating prototypes
● While there are some cons, we have been able to handle them
so far, so overall, we believe it was a good choice to adopt it
Conclusion
Slide 39
Slide 39 text
THANK YOU