when we made the change to async persist notifications, we forgot to
pass through api_key_id and key_type. in send_sms/email, for legacy
reasons, they default to None/KEY_TYPE_NORMAL, so regardless of what
your api key was set up as, we would send real messages!
TODO: Once the PaaS transition is complete and the task changes are
reverted, remove the api_key_id and key_type params from the send_*
tasks entirely, as those are only called from the csv job flow, and
don't need them
This is being done for the PaaS migration to allow us to keep traffic coming in whilst we migrate the database.
uses the same tasks as the CSV uploaded notifications. Simple changes to not persist the notification, and call into a different task.
There are three authentication methods:
- requires_no_auth - public endpoint that does not require an Authorisation header
- requires_auth - public endpoints that need an API key in the Authorisation header
- requires_admin_auth - private endpoint that requires an Authorisation header which contains the API key for the defined as the client admin user
> If a user makes an API request with additional personalisation fields,
> we should simply discard any fields that the template doesn't have.
>
> This gives a couple of related advantages:
>
> - modifying template parameters no longer requires downtime for
> clients - as they can pass in extra new parameters before a template
> change, or continue passing in old unused parameters after removing
> them from a template
>
> - services can pass in large user objects, for example, and then play
> around with templates adding and removing fields at will
>
> we should make sure we still return an error if a user doesn't pass in
> a required parameter.
– https://www.pivotaltracker.com/story/show/140774195
This PR changes the response to POST /notifications/sms/<mmg | firetext> from a 400 response to a 200 response.
If we get a callback for a notification more than once or for a notification we log that but we return a 200 success response to the provider.
We have found that there is a situation where the send to provider throws a timeout exception but the provider did get the message, but we still send it to them again.
In which case they send the message twice, and callback for the message twice.
Another case where we may get duplicate callbacks is that the network gave the provider two callbacks meaning they pass those two callbacks onto us.
So it is really difficult to know if we sent to the provider twice or just got two callbacks.
The test_callback has many changes because I took the opportunity to use the client conftest fixture rather than the notify_api fixture.
The only 2 tests really changed are test_mmg_callback_returns_200_when_notification_id_not_found_or_already_updated and test_firetext_callback_returns_200_when_notification_id_not_found_or_already_updated
We are using the notify queue in this iteration because that queue is a low volume queue with it's own dedicated workers. This just saves us from building a new queue at this point, and a new queue may not be necessary.
- Added the `simulate` notification logic to version 2. We have 3 email addresses and phone numbers that are used
to simulate a successful post to /notifications. This was missed out of the version 2 endpoint.
- Added a test to template_dao to check for the default value of normal for new templates
- in v2 get_notifications, casted the path param to a uuid, if not uuid abort(404)
- note this is an unexpectedly big change.
- When we create a service we pass the service id to the persist method. This means that we don't have the service available to check if in research mode.
- All calling methods (expecting the one where we use the notify service) have the service available. So rather than reload it I changed the method signature to pass the service, not the ID to persist.
- Touches a few places.
Note this means that the update or create methods will fall over on a null service. But this seems correct.
Goes back to the story which we need to play to make the service available as the API user so that the need to load and pass around services is minimised.
1) It's incr not inc on the redis client, so renamed the calls everywhere
2) Redis returns bytes/string rather than an int if the value stored is an int. Cast the result to an int before use. Not you can set up the GET to do this transparently but I've not done this as we *may * use GETS for non-int and the callback sets up the cast for the connection not the call.
These means that the cache count is on Notifications in the database NOT notifications sent to providers. If the provider fails to accept the notification, it still counts.
I think this is correct, as they have done the work to send it so we should count it, though there is an argument that we should count them on sending?
- Uses Redis cache to check for current count
- If not present then sets the value based on the database state
- Any Redis errors are swallowed. Cache failures should NOT fail the request.
- It would be nice to refactor the send_sms and send_email tasks to use these common functions as well, that way I can get rid of the new Notifications.from_v2_api_request method.
- Still not happy with the format of the errors. Would like to find a happy place, where the message is descript enough that we do not need external documentation to explain the error. Perhaps we still only need documentation to explain the trial mode concept.
- Use these validation methods in post_sms_notification and the version 1 of post_notification.
- Create a v2 error handlers.
- InvalidRequest has a to_dict method for private and v1 error responses and a to_dict_v2 method to create the v2 of the error responses.
- Each validation method has extensive unit tests, so the unit test for the endpoint do not need to check every error case, but check that the error handle formats the message correctly.
- The format of the error messages is still a work on progress.
- This version of the api could be deployed without causing a problem to the application.
- The new endpoing is still a work in progress and is not being used yet.
when given any log function with multiple parameters, the python logging utils
assume the first param is a format string and the rest are arguments to pass
in - we were passing in the exception object to `logger.exception`, however,
the purpose of .exception is to add the exception object itself - so we didn't
need to
Currently getting a single notification by ID is restricted to
notifications created with the same key type.
This makes things awkward for the functional tests now we’ve removed the
ability to create live keys in trial mode. So this commit removes the
restriction, so that any key can get any notification, no matter how it
was created.
And you’re never going to guess a UUID, so the chances of this giving
you privileged access to someone’s personal information is none.
This does not change the get all notifications endpoint, which
absolutely should be restricted by key type.
Refactored send_notifications method so that it is more readible.
Refectored the test_send_notificaitons so that it uses parametrized test to avoid duplication.
