Python/Django support for distributed multi-tenant databases like Postgres+Citus
Enables easy scale-out by adding the tenant context to your queries, enabling the database (e.g. Citus) to efficiently route queries to the right database node.
There are architecures for building multi-tenant databases viz. Create one database per tenant, Create one schema per tenant and Have all tenants share the same table(s). This library is based on the 3rd design i.e Have all tenants share the same table(s), it assumes that all the tenant relates models/tables have a tenant_id column for representing a tenant.
The following link talks more about the trade-offs on when and how to choose the right architecture for your multi-tenat database:
https://www.citusdata.com/blog/2016/10/03/designing-your-saas-database-for-high-scalability/
Other useful links on multi-tenancy:
- https://www.citusdata.com/blog/2017/03/09/multi-tenant-sharding-tutorial/
- https://www.citusdata.com/blog/2017/06/02/scaling-complex-sql-transactions/
pip install --no-cache-dir django_multitenant
Python | Django |
---|---|
3.X | 2.2 |
3.X | 3.2 |
3.X | 4.0 |
In order to use this library you can either use Mixins or have your models inherit from our custom model class.
- In whichever files you want to use the library import it:
from django_multitenant.fields import * from django_multitenant.models import *
- All models should inherit the TenantModel class.
Ex: class Product(TenantModel):
- Define a static variable named tenant_id and specify the tenant column using this variable.
Ex: tenant_id='store_id'
- All foreign keys to TenantModel subclasses should use TenantForeignKey in place of models.ForeignKey
- A sample model implementing the above 2 steps:
class Store(TenantModel):
tenant_id = 'id'
name = models.CharField(max_length=50)
address = models.CharField(max_length=255)
email = models.CharField(max_length=50)
class Product(TenantModel):
store = models.ForeignKey(Store)
tenant_id='store_id'
name = models.CharField(max_length=255)
description = models.TextField()
class Meta(object):
unique_together = ["id", "store"]
class Purchase(TenantModel):
store = models.ForeignKey(Store)
tenant_id='store_id'
product_purchased = TenantForeignKey(Product)
- In whichever files you want to use the library import it by just saying
from django_multitenant.mixins import *
- All models should use the
TenantModelMixin
and the djangomodels.Model
or your customer Model classEx: class Product(TenantModelMixin, models.Model):
- Define a static variable named tenant_id and specify the tenant column using this variable.
Ex: tenant_id='store_id'
- All foreign keys to TenantModel subclasses should use TenantForeignKey in place of models.ForeignKey
- Referenced table in TenenatForeignKey should include a unique key including tenant_id and primary key
Ex: class Meta(object): unique_together = ["id", "store"]
- A sample model implementing the above 3 steps:
class ProductManager(TenantManagerMixin, models.Manager):
pass
class Product(TenantModelMixin, models.Model):
store = models.ForeignKey(Store)
tenant_id='store_id'
name = models.CharField(max_length=255)
description = models.TextField()
objects = ProductManager()
class Meta(object):
unique_together = ["id", "store"]
class PurchaseManager(TenantManagerMixin, models.Manager):
pass
class Purchase(TenantModelMixin, models.Model):
store = models.ForeignKey(Store)
tenant_id='store_id'
product_purchased = TenantForeignKey(Product)
objects = PurchaseManager()
- Creating foreign keys between tenant related models using TenantForeignKey would automate adding tenant_id to reference queries (ex. product.purchases) and join queries (ex. product__name). If you want to ensure to create composite foreign keys (with tenant_id) at the db layer, you should change the database ENGINE in the settings.py to
django_multitenant.backends.postgresql
.
'default': {
'ENGINE': 'django_multitenant.backends.postgresql',
......
......
......
}
-
Write authentication logic using a middleware which also sets/unsets a tenant for each session/request. This way developers need not worry about setting a tenant on a per view basis. Just set it while authentication and the library would ensure the rest (adding tenant_id filters to the queries). A sample implementation of the above is as follows:
from django_multitenant.utils import set_current_tenant class MultitenantMiddleware: def __init__(self, get_response): self.get_response = get_response def __call__(self, request): if request.user and not request.user.is_anonymous: set_current_tenant(request.user.employee.company) return self.get_response(request)
In your settings, you will need to update the
MIDDLEWARE
setting to include the one you created.MIDDLEWARE = [ # ... # existing items # ... 'appname.middleware.MultitenantMiddleware' ]
-
Set the tenant using set_current_tenant(t) api in all the views which you want to be scoped based on tenant. This would scope all the django API calls automatically(without specifying explicit filters) to a single tenant. If the current_tenant is not set, then the default/native API without tenant scoping is used.
def application_function: # current_tenant can be stored as a SESSION variable when a user logs in. # This should be done by the app t = current_tenant #set the tenant set_current_tenant(t); #Django ORM API calls; #Command 1; #Command 2; #Command 3; #Command 4; #Command 5;
- Most of the APIs under Model.objects.*.
- Model.save() injects tenant_id for tenant inherited models.
s=Store.objects.all()[0]
set_current_tenant(s)
#All the below API calls would add suitable tenant filters.
#Simple get_queryset()
Product.objects.get_queryset()
#Simple join
Purchase.objects.filter(id=1).filter(store__name='The Awesome Store').filter(product__description='All products are awesome')
#Update
Purchase.objects.filter(id=1).update(id=1)
#Save
p=Product(8,1,'Awesome Shoe','These shoes are awesome')
p.save()
#Simple aggregates
Product.objects.count()
Product.objects.filter(store__name='The Awesome Store').count()
#Subqueries
Product.objects.filter(name='Awesome Shoe');
Purchase.objects.filter(product__in=p);
This library uses similar logic of setting/getting tenant object as in django-simple-multitenant. We thank the authors for their efforts.
Copyright (C) 2018, Citus Data Licensed under the MIT license, see LICENSE file for details.