diff --git a/docs/graphql/manual/remote-schemas/index.rst b/docs/graphql/manual/remote-schemas/index.rst index 147c03e0794..a753a694d5c 100644 --- a/docs/graphql/manual/remote-schemas/index.rst +++ b/docs/graphql/manual/remote-schemas/index.rst @@ -6,10 +6,13 @@ Remote schemas :depth: 2 :local: -Hasura gives you CRUD + realtime GraphQL APIs with authorization & access control. However, in many cases, you will need to write APIs (queries, mutations) that contain custom logic. For example, implementing a payment API, or querying data that is not in your database. +Hasura gives you CRUD + realtime GraphQL APIs with authorization & access control. However, in many cases, you will +need to write APIs (queries, mutations) that contain custom logic. For example, implementing a payment API, or +querying data that is not in your database. Hasura has the ability to merge remote GraphQL schemas and provide a unified GraphQL API. Think of it -like automated schema stitching. All you need to do is build your own GraphQL service and then provide the HTTP endpoint to Hasura. Your GraphQL service can be written in any language or framework. +like automated schema stitching. All you need to do is build your own GraphQL service and then provide the HTTP +endpoint to Hasura. Your GraphQL service can be written in any language or framework. This is what Hasura running with "Remote schemas" looks like: @@ -20,7 +23,7 @@ This is what Hasura running with "Remote schemas" looks like: .. note:: - This is a new feature in active development. Please do give us feedback, bug-reports and ask us questions on + This is a new feature under active development. Please do give us feedback, bug-reports and ask us questions on our `discord `__ or on `github `__. Use-cases @@ -31,49 +34,38 @@ Use-cases You can handle these use-cases by writing resolvers in a custom GraphQL server -and making Hasura merge this ``remote schema`` with the existing autogenerated +and making Hasura merge this "remote schema" with the existing auto-generated schema. You can also add multiple remote schemas. Think of the merged schema as a union of top-level nodes from each of the sub-schemas. -Note that if you are looking for adding authorization & access control for your -app users to the GraphQL APIs that are auto-generated via Hasura, head to -:doc:`Authorization / Access control <../auth/index>` +.. note:: -.. admonition:: Nomenclature + If you are looking for adding authorization & access control for your + app users to the GraphQL APIs that are auto-generated via Hasura, head to + :doc:`Authorization / Access control <../auth/index>` - Top-level node names need to be unique across all merged schemas (*case-sensitive match*). - Types with the *exact same name and structure* will be merged. But types with the *same name but different structure* will result in type conflicts. +Adding a remote schema +---------------------- +Follow the steps below to add a "remote schema" to Hasura GraphQL engine: -How to add a remote schema --------------------------- - -Follow the steps below to add your "remote schema" to hasura. - -Step-1: Write a custom GraphQL server +Step 1: Write a custom GraphQL server ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ You need to create a custom GraphQL server with a schema and corresponding resolvers that solve your use case -(*if you already have a functional GraphQL server that meets your requirements, you can skip this step*). You can -use any language/framework of your choice to author this server or deploy it anywhere. A great way to get started -is to use one of our boilerplates: +(*if you already have a functional GraphQL server that meets your requirements, you can skip this step*). + +You can use any language/framework of your choice to author this server and deploy it anywhere. A great way to get +started is to use one of our boilerplates: - `Boilerplates `__ - -.. admonition:: Current limitations - - - Nodes from different GraphQL servers cannot be used in the same query/mutation. All top-level nodes have to be from the same GraphQL server. - - Subscriptions on remote GraphQL server are not supported. - - Interfaces_ and Unions_ are not supported - if a remote schema has interfaces/unions, an error will be thrown if you try to merge it. - - These limitations will be addressed in upcoming versions. - -Step-2: Merge remote schema +Step 2: Merge remote schema ^^^^^^^^^^^^^^^^^^^^^^^^^^^ -Head to the console to merge your remote schema with GraphQL Engine's auto-generated schema. In a top level tab, -named ``Remote Schemas``, click on the ``Add`` button. +To merge your remote schema with GraphQL Engine's auto-generated schema: + +Head to the ``Remote Schemas`` tab of the console and click on the ``Add`` button .. image:: ../../../img/graphql/manual/business-logic/add-remote-schemas-interface.png @@ -89,40 +81,120 @@ You need to enter the following information: - **Headers**: configure the headers to be sent to your custom GraphQL server. - Toggle forwarding all headers sent by the client (when making a GraphQL query) to your remote GraphQL server. - - Send additional headers to your remote server - These can be static header name-value pairs; and/or pairs of "header name-environment variable name". - You can specify the value of the header to picked up from the enviroment variable. + - Send additional headers to your remote server - These can be static header name-value pairs; and/or pairs of + "header name-environment variable name". You can specify the value of the header to picked up from the environment + variable. - **Example**: Let's say your remote GraphQL server needs a ``X-Api-Key`` as a header. As this value contains sensitive data (like API key in this - example), you can configure name of an environment variable which will hold the value. This environment variable needs to be present when you start - GraphQL Engine. When Hasura sends requests to your remote server, it will pick up the value from this environment variable. + **Example**: Let's say your remote GraphQL server needs a ``X-Api-Key`` as a header. As this value contains + sensitive data (like API key in this example), you can configure name of an environment variable which will hold + the value. This environment variable needs to be present when you start GraphQL Engine. When Hasura sends + requests to your remote server, it will pick up the value from this environment variable. -.. note:: +.. admonition:: Using environment variables - If the remote schema configuration contains environment variables - either - for URL or headers - **environment variables need to be present** (GraphQL - engine should be started with these env variables) with valid values, when - adding the remote schema. + If you are using environment variables in the remote schema configuration - either + for URL or headers - **the environment variables need to be present** with valid values + when adding the remote schema i.e. GraphQL engine should be started with these env variables Click on the ``Add Remote Schema`` button to merge the remote schema. - -Step-3: Make queries to the remote server from Hasura +Step 3: Make queries to the remote server from Hasura ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ -Now you can head to *GraphiQL* and make queries to your remote server from Hasura. +Now you can head to the ``GraphiQL`` tab and make queries to your remote server from Hasura. -Query your remote server by making requests to the Hasura graphql endpoint (``/v1alpha1/graphql``). +You can query your remote server by making requests to the Hasura GraphQL endpoint (``/v1alpha1/graphql``). + +Points to remember +------------------ + +Remote schema fields nomenclature +^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ + +- Top-level field names need to be unique across all merged schemas (*case-sensitive match*). +- Types with the *exact same name and structure* will be merged. But types with the *same name but different + structure* will result in type conflicts. -.. note:: +Current limitations +^^^^^^^^^^^^^^^^^^^ - For some use cases, you may need to extend the GraphQL schema fields exposed by Hasura GraphQL engine - (*and not merely augment as we have done above*) with a custom schema/server. To support them, you can use - community tooling to write your own client-facing GraphQL gateway that interacts with GraphQL Engine. - - But adding an additional layer on top of Hasura GraphQL engine significantly impacts the performance provided by it - out of the box (*by as much as 4x*). If you need any help with remodeling these kind of use cases to use the - built-in remote schemas feature, please get in touch with us on `Discord `__. +- Nodes from different GraphQL servers cannot be used in the same query/mutation. All top-level fields have to be + from the same GraphQL server. +- Subscriptions on remote GraphQL servers are not supported. +- Interfaces_ and Unions_ are not supported - if a remote schema has interfaces/unions, an error will be thrown if + you try to merge it. +These limitations will be addressed in upcoming versions. .. _Interfaces: https://graphql.github.io/learn/schema/#interfaces .. _Unions: https://graphql.github.io/learn/schema/#union-types + +Extending the auto-generated GraphQL schema fields +^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ + +For some use cases, you may need to extend the GraphQL schema fields exposed by Hasura GraphQL engine +(*and not merely augment as we have done above*) with a custom schema/server. To support them, you can use +community tooling to write your own client-facing GraphQL gateway that interacts with GraphQL Engine. + +.. note:: + + **Adding an additional layer on top of Hasura GraphQL engine significantly impacts the performance provided by + it out of the box** (*by as much as 4x*). If you need any help with remodeling these kind of use cases to use the + built-in remote schemas feature, please get in touch with us on `Discord `__. + +Bypassing Hasura's authorization system for remote schema queries +----------------------------------------------------------------- + +It might be necessary sometimes to bypass Hasura's authorization system (calling +the configured webhook, or validating the JWT), for queries that are for a +remote GraphQL server. + +**For example**, you have a remote GraphQL server which does authentication, +i.e. signup and login, and you have added it as a remote schema. In this case, +you would not want to perform Hasura's authorization when the user is making a +login/signup request. + +There is no first-class option to currently do this via any configuration in +Hasura. However a similar solution can achieved by the following workarounds: + +Bypassing webhook authorization +^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ + +If you have a :doc:`webhook authorization setup <../auth/webhook>`, in the normal scenario, your authorization +webhook would return ``200`` on success and ``401`` if it is either unable to authorize the current request or if +the authorization information is absent (like cookie, authorization header etc.) + +To bypass the webhook auth: + +- the webhook should respond with ``200`` and ``x-hasura-role: anonymous`` instead of a ``401`` when the + authorization information is absent or if it fails to resolve the authorization information. +- when adding the remote schema, check the ``Forward all headers from client`` option so that the remote server + will get the relevant cookie/header (from the client) and the role ``anonymous``. + +Bypassing JWT authorization +^^^^^^^^^^^^^^^^^^^^^^^^^^^ + +If you have a :doc:`JWT authorization setup <../auth/jwt>`, to bypass the JWT auth: + +- your authentication server should generate a static JWT token for ``anonymous`` i.e. unauthenticated users. +- when adding the remote schema, check the ``Forward all headers from client`` option so that the remote server + will get the JWT (from the client). + +For example, the generated JWT can be: + +.. code-block:: json + + { + "sub": "0000000000", + "iat": 1516239022, + "role": "anonymous", + "https://hasura.io/jwt/claims": { + "x-hasura-allowed-roles": ["anonymous"], + "x-hasura-default-role": "anonymous" + } + } + + +Hasura will get this JWT and successfully validate it. When your remote server receives this JWT, it should +specifically validate the JWT and, for example, check for ``role`` key in the JWT. If it is set to ``anonymous`` +then it should consider the request as unauthenticated.