Note that this guide is for the main Tray platform. If you are an Embedded customer please see the Error Handling guide for Tray Embedded edition
If any of the steps in your workflow encounter errors when communicating with third-party services (e.g. your Opsgenie API key doesn't have the correct access rights or you are trying to access a record that does not exist in Salesforce, Marketo etc.), you can choose how you would like to handle these errors.
The error handling methods explained below only concern error messages returned by third-party service APIs when operations are communicating with them. If misconfigured, some operations will fail in the Tray workflow builder before they communicate with the service.
For example, if you are using the Slack Send message operation and have changed the channel drop-down to a jsonpath but mis-form your jsonpath (e.g. by entering
steps.trigger.data instead of
$.steps.trigger.data) then the connector will not actually send a request to the Slack service API in order to receive a recognized error message.
To illustrate this, in the following workflow section the Set slack name in list data storage step might fail your workflow because it uses the
$.steps.sheets-1.results.result} jsonpath to get a name which was looked up in Google Sheets, and Sheets could return an empty result:
The problem here is that if Google Sheets returns an empty result, it is not considered an error by the Google Sheets API - it is just informing you that the result is empty (note that some services will return an error for empty results - it depends on how the service API works!).
Subsequently, when the Set slack name in list step attempts to retrieve the result you will get the following Tray system error:
"message": "Reference: $.steps.sheets-1.results.result in property: 'value' did not resolve to any value."
Because this is not an API Error message returned by a third party, you will not be able to use the continue / manual / stop methods discussed below.
So we deal with this by setting a boolean step to check if the Google Sheets step has returned an empty array:
We know that we needed to check for an empty array because inspection of the logs on the left shows that the result can be:
If the result of this is true (i.e. it is not an empty array) then we can set the name in the list.
If it is false, then we take no action and just continue the loop to process the next name in the list.
For these types of errors, this is how we can set up an equivalent of the 'continue' error option as described below. For your own use case, you may wish to put extra steps in the false branch, to create an equivalent of the 'manual' option described below.
Errors are only sent to alerting workflows once all retries of a workflow have failed.
Multiple workflows can use the same Alerting Workflow but a workflow can't send its errors to more than one Alerting Workflow.
There are 3 ways of dealing with errors:
Continue - if the error is insignificant then you can continue the workflow as normal
Manual - it is possible to create 'Success' and 'Error' branches coming from the connector step in question, so that you can set exactly what happens in the workflow should an error be encountered with this step. This works much the same as a boolean connector
Stop workflow (default) - in this case the workflow is stopped and an error payload can be sent to an error handling workflow which you need to set up
Choosing the Error Handling method
For any connector step, you can find the Error Handling method drop-down by scrolling down to the very bottom of the operations panel:
The Continue workflow option
If you select the Continue option the workflow will continue on to the next step, but you will notice that the step in question has a blue alert symbol next to it:
Note that you should be careful about using the continue method as, if other steps are dependent on any output data from it, it could break your workflow.
The Manual option
If you choose Manual Alerting you will see a green success and red failure branch coming from your connector step. So you can then, for example, send an alerting email (this is of course a very simple example, and you could use the error branch to build a sophisticated alternative route for your workflow):
The above example shows how to compile a simple email (this could also be a message to a Slack channel or any other option you might want to implement) which makes use of the error message from the output of the errored step.
Accessing error messages
Looking at the debug output of an errored step shows you what fields are available:
Note: when referencing the error message from the step, it must be done in the format
$.errors.<step-name>.message. The above example makes use of interpolated mode by inserting the error into the content of the email message with
This then picks up any API errors that might be generated and sends them in a composed email such as:
The Stop Workflow (Default) option
Errors can be sent to another workflow and that workflow is responsible for handling the alerting. One case for this would be to use the HTTP connector to pass on any errors to your monitoring system of choice.
There are two steps to setting this up:
- Create an alerting workflow using the Alerting Trigger
- Configure your workflows to send their errors to your alerting workflow
You can use the same alerting workflow to handle the errors from multiple workflows. For example, you might set up your alerting workflow to send you an email with details of each error it receives. You can set all of your existing workflows to report their errors in this way.
Creating an alerting workflow
Create a new workflow and select the Alerting Trigger as its trigger.
This is a special kind of trigger that can accept alerting payloads from other workflows. Alerting payloads are in the following format:
|workflow_uuid||The UUID of the workflow which sent the error|
|workflow_url||The URL of the workflow which sent the error|
|workflow_title||The title of the workflow which sent the error|
|step_name||The name of the specific step which was the cause of the error. Note that the is not the user-configured name but the unique name given by Tray|
|step_log_url||This URL can be used to go directly to the point in the logs at which this error occurred|
|connector||The type of connector (and version) that caused the error|
|message||The internal Tray-generated error message|
|created||A timestamp of when the error occurred|
Your alerting workflow can make use of these fields to provide useful information for debugging. The payload includes the uuid, url, and title of the workflow which has errored as well as the specific step name and a link to the step in logs. This link to the logs is useful for debugging the cause of the error.
Workflows with an alerting trigger behave just like any other workflow. You can build any logic you like into your workflow to handle the incoming errors.
Configuring your workflow to send errors to your alerting workflow
Click on the cog in the top left corner of the builder. This will open the Workflow Settings panel. From here you can set the alerting workflow which will handle the errors from this workflow (you may need to refresh the page to see a just-created alerting workflow in the list). Note that only workflows with an Alerting Trigger will appear in this list.