Responding to asynchronous events

Event notifications are primarily used for one-way, event-based data streams where systems SEND or RECEIVE data asynchronously. You can think of these like CRUD type of requests you want to send to or receive from your connection's EHR system.

Recoverable messages

We use recoverable messages to guarantee delivery of asynchronous traffic. We save these messages into highly durable queues so that they can survive restarts or other failures. We won't respond with a 200 OK status unless we successfully saved the message in a highly durable queue first.

Read about our business continuity and disaster recovery plan.

Sequencing asynchronous events

Event notifications are delivered via Redox to the desired system using a first-in, first-out (FIFO) method. We work to maintain the order of asynchronous events in the order they're triggered.

One exception to that, though, is that Redox sometimes processes events in parallel before sending in FIFO order. So you shouldn't use the log ID to sequence requests, especially if the endpoint receives different data models or FHIR® resources. Instead, use the timestamp for when the data arrived at the endpoint to sequence it against other asynchronous traffic.

If failures happen during processing, we pause the queue for the intended endpoint so that we can maintain FIFO delivery after the issue is resolved.

Meanwhile, we send a notification to the sender or receiver, depending on where the failure occurred. Either way, the notification goes to the Alert Email, which is listed in the Organization Profile of the Redox dashboard.

New beta option: Child queues for subscriptions

Are you a little worried about the high volume of async traffic you might have? If so, you may want to think about using child queues for your data model subscriptions.

Child queues allow you to have multiple queues for one subscription so that messages process in parallel. Using this option can speed up your throughput, so long as the destination can handle that volume of messages.

Learn more about child queues for subscriptions and talk to your Redox account manager about setting up child queues for one of your data model subscriptions.

Responses to SEND

Some packages don’t get to their intended destination. Amazon knows this well.

Sometimes your asynchronous messages also don’t get to their intended endpoint(s). The good news is that we can let you know whether your message arrived safely. But remember that asynchronous messages are one-way, so you won’t get a response back from your connection's system. Much like Amazon doesn’t get thank-you packages for successfully delivered orders.

Any responses to asynchronous traffic are from Redox specifically about your outgoing message. And if your message went to multiple endpoints, you receive only one overall response. Check out the details about each scenario below.

Successful response

If Redox successfully accepts your message, you receive a 200 OK status and a response body with any relevant details.

A successful status doesn't mean that the message was delivered to the intended endpoint. It only means that we received the message, saved it to a highly durable queue, and are ready to process. Then, we'll deliver it to the intended endpoint(s).

Later, if you want to know for certain whether your message was successfully processed and delivered, you can check the status of the related log. Learn more about logs.

The difference between a response and a log is that Redox sends a response to a message for multiple endpoints, but we create a log for each endpoint. So if you send one message to 10 endpoints, you receive one response with 10 log objects and IDs in Meta.Logs[].ID array (for the Data Model API).

Unsuccessful response

On occasion, Redox can’t process a request because of an issue with the request or within Redox itself. If this happens, you receive a 4XX or 5XX HTTP status code and a response with the array in the body populated: Meta.Errors[] array (for the Data Model API) or the issue[] array (for the FHIR® API). These arrays contain an error object with the type of failure and corresponding details. At a minimum, a Text field provides the error message details.

Example: Unsuccessful SEND response for the Data Model API

1
{
2
   "Meta": {
3
     …
4
     "Errors": [
5
       {
6
         "Text": "Required field missing - Device.ID in Device:New"
7
         …
8
       }
9
     ]
10
   }
11
   …
12
 }

Example: Unsuccessful SEND response for the FHIR API

1
{
2
  "resourceType": "OperationOutcome",
3
  "issue": [{
4
    "severity": "error",
5
    "code": "invalid",
6
    "details": {
7
      "text": "Unknown search parameter name \"foobar\" for model \"Patient\"",
8
      "coding": [
9
        {
10
          "code": "MSG_PARAM_UNKNOWN",
11
          "system": "http://terminology.hl7.org/CodeSystem/operation-outcome",
12
          "display": "Parameter \"foobar\" not understood"
13
        }
14
      ]
15
    },
16
    "diagnostics": "Failed to execute FHIR search"
17
  }]
18
}

As for the HTTP status codes, you may see any of these:

Code	Description
`400 Bad Request`	The request is formatted incorrectly or is missing required information.
`401 Unauthorized`	Authentication failed. This means you need to refresh or regenerate your access token and try again.
`403 Forbidden` or `404 Not Found`	The request is missing the (a) `POST` verb; (b) designated endpoint; or (c) specified data model/resource after `api.redoxengine.com` (or `api.ca.redoxengine.com` for Canada).
`413 Payload Too Large`	The payload sent exceeds the payload size limit of 10 MB. Refer to Redox message size limits.
`429 Too Many Requests`	Your system has exceeded the rate limit of 7,500 requests per IP address for the current five-minute rolling window. Wait a few minutes then try initiating the request again.
`500 Internal Server Error`	An unknown error occurred within Redox.
`502 Bad Gateway`	The request was dropped within Redox. This can sometimes occur during brief periods when internal updates are being applied.
`504 Gateway Timeout`	The request took too long to complete. Requests to Redox time out after a couple of minutes.

We recommend that your system have queuing capability for your outgoing requests to Redox in case of any kind of failure. This way, your system can queue any additional outgoing requests until the issue is resolved.

For any of the 5XX errors, you can retry sending the request to see if the Redox issue is resolved. We recommend using an exponential backoff strategy between retries. If the error persists though, submit a ticket via our Help Desk.

Responses to RECEIVE

We rely on you to send an HTTP status code to let us know whether you successfully received a message from your connection's system. We wait for your response before deciding what to do next.

Successful receipt

If you successfully receive an event from Redox, your system should send back a 200 OK status with an empty Meta.Errors[] array (for the Redox Data Model API) or issue[] array (for the FHIR® API) in the body of the response. This response means that your system has processed the most recent data and is ready for the next payload. We proceed with sending your next asynchronous data payload.

Unsuccessful receipt

If unable to receive an event, your system should send back to Redox the relevant 4XX or 5XX HTTP status code and a response with the relevant array in the body populated: Meta.Errors[] array (for the Redox Data Model API) or the issue[] array (for the Redox FHIR® API). This array should contain an error object with the type of failure and corresponding details. At a minimum, a Text field should provide the error message details.

What your connection doesn't know

The failed response comes to us, but we don't pass it back to your connection's system. Your connection won't know that your system failed to receive their event notification.

It's possible to set up custom notifications as an advanced feature of Redox Nexus. Talk to a Redoxer to find out more.

If you send us a failed HTTP status code in response to an asynchronous message, our next action depends on the type of environment your endpoint is in.

For a production environment, we briefly retry sending the message after a brief pause then pause the queue.

For a staging or development environment, we retry sending the message after a brief pause. But we don’t pause the queue of asynchronous messages from your connection's system.

Non-critical errors

For non-critical errors, we expect that you send back a 200 OK status prior to performing data validation or after finding an error that can be resolved at a later time. This way, we don’t pause your data for production environments and your development/staging environments can proceed with business as usual. This option ensures continuity in your data delivery so that your data flow doesn’t get hung up on a single troublesome data payload.

If you do identify a non-critical issue with an incoming request, we recommend that you log an error in your system and submit a ticket via our Help Desk if you need help resolving it.

Alternatively, if you're experiencing downtime or another critical error, we recommend that you send a 4XX or 5XX status to initiate pausing on our end so that none of your data is lost while you work to resolve your system.

Automatic retries

The process for automatic retries occurs differently based on the environment type.

Production environments

If an asynchronous message fails in a production environment, we retry the failed message until it succeeds. We also pause the queue for subsequent messages, so traffic is paused until the failed one successfully goes through. Our Integration team receives notifications if any message queues pile up so they can review and resolve any processing failures.

Failed test requests for production

We ignore delivery failures for test requests. This means we don’t automatically retry or pause the queue when test requests fail sending to a production environment.

But let’s explain how we retry your failed request with that analogy of a mail carrier delivering a physical package:

The mail carrier rings the doorbell the first time, but no one answers. They wait a few seconds, then try ringing the doorbell again. There’s still no answer, but this is a determined—but polite—mail carrier. They wait a little longer before retrying a third time. Even then, the mail carrier doesn’t give up. They keep ringing the doorbell, but they wait a little bit longer between each try, so as to not be obnoxious while being persistent. Finally, the door opens, the receiver accepts the package, and the mail carrier continues on to deliver the next package.

In Redox terms, we follow the same pattern: waiting a few seconds before retrying, then steadily increasing the wait period until it becomes a fixed wait of a minute or two between requests. Retries continue every couple of minutes until the receiving endpoint acknowledges they received the data, or until we manually pause the request.

Unexpected pauses

Retries may stop unexpectedly if we receive a different error code from the recipient system that initiated the request. We review these types of alerts daily to see if we need to manually intervene. If so, we reach out to the recipient's team to resolve issues.

Custom notifications

We're all about that "never surrender" life, but we do let you know if we're having trouble delivering a request after our third failed attempt. We notify you via the email address listed in your alert email, which is in the Organization Profile page of the dashboard. Make sure that email address is up-to-date so that we can let you know when there's trouble in paradise—and when we've found paradise again by successfully delivering the request.

If you would like, we can set up more custom notifications for you, whether via email or Slack. For example, you can receive notifications for each endpoint that fails to accept one of your requests. Submit a ticket via our Help Desk to talk about your notification options.

Development or staging environments

If an asynchronous message fails in a development or staging environment, we don't automatically retry the failed message or pause the queue. This means that other asynchronous messages can continue sending.

Failed test requests for development or staging

We ignore delivery failures for test requests. This means we don’t automatically retry when test requests fail sending to a development or staging environment.

Manual retries

FHIR® is a registered trademark of Health Level Seven International (HL7) and is used with the permission of HL7. Use of this trademark does not constitute an endorsement of products/services by HL7®.