Alerts & response parsing using script

[leevik]

One of my integrations destination API returns error messages that requires some digging. I have used following structure in destinations response transformer:
if(msg.hasOwnProperty("error")){
var customError = {
"errorType": "Client",
"Guid": $('clientGuid'),
"error": msg
}
router.routeMessageByChannelId('example', JSON.stringify(customError));
}

I am wondering if I could use some postprocessor script or something to route error messages to channel I want since it is bit tedious to copy paste in every single destination's response transformer. Something to affect whole channel would be great.

Also I get errors as HTTP Sender error, are these destination connector errors or what?

[pacmano1]

I'm not sure what you mean by tedious, since it's a literally a four second thing to copy and paste.

I'd also question why you'd be cutting and pasting it so many times, since if you build channels that are multitenant capable, you simply don't have to do that.

Firstly, in order for a response transformer to always fire its data type must be set to raw. Typically, you then look at the http status code and proceed accordingly for your use case. That includes possibly requeuing the message and/or eventually setting the message to an error. It also means you need to specifically parse the response body.

You could use a global post processor, but I suspect you might want to specifically handle errors in different ways based on what end point was called.

[leevik]

Let's say I have 10 clients that use same API. All 10 clients have 10 channels and all 10 channels have 5 destinations. It is 500 copy paste even with that few clients.
Here is an example of an error.

{
  "error" : {
    "httpStatusCode" : 400,
    "type" : "InvalidOperation",
    "details" : [
      {
        "message" : "Contact already exist for this account"
      }
    ]
  }
}

My clients can't do anything with that error message so I have to enrich it.
That is why I use for example something like this depending on the channel (on every destination response transformer):

channelMap.put("name", msg.contact.name)
if(msg.hasOwnProperty("error")){
	var customError = {
		"errorType": "Contact",
		"Guid": channelMap.get("name"),
		"errorMessage":msg
	}
	router.routeMessageByChannelId('12345', JSON.stringify(customError));

Then on the error handling channel I do:

var customError = JSON.parse(msg)
if(customError.errorType == "Contact"){
  alert.sendAlert("Contact: " + customError.Guid + " has an error on update/add: " + customError.errorMessage)
}

I could do this kind of thing on the response transformer, but I thought this was better for keeping up with the errors. But if I could do that "msg.hasOwnProperty("error")" route to channel 12345 per channel not per destination it would be better.

[pacmano1]

I would not have 10 clients calling 10 different endpoints for the same API set. That is not a scalable. Instead have all 10 clients calling the same set of API's delivered on a single channel. For more capacity you can multi thread the channel and/or clone the channel. The single http listener can route to other channels and return the response back upstream to the original calling channel. Or set up a channel per API endpoint and use a load balancer to direct based on the uri. Naturally, you'd typically have a load balancer in front of many solutions. There are lots of ways to do this other than what appears to be your approach if I'm understanding it correctly.

Obviously, I'm suggesting there's something fundamentally wrong with how you're designing your channels.

You could try to do something in the Global Post Processor, but it would apply to all channels - so that code would likely need to evaluate "this channel name" to determing if in fact it should do anything.