missing cluster error handling

[eguzki]

When the service endpoint does not exist from the defined set of envoy clusters.

"services": {
  "mistyped-service": {
    "type": "ratelimit",
    "endpoint": "does-not-exist",
    "failureMode": "deny"
  }
},
"actionSets": [
{
    "actions": [
        {
            "service": "mistyped-service",
            "scope": "b",
            "data": [
                {
                    "expression": {
                        "key": "limit_to_be_activated",
                        "value": "1"
                    }
                }
            ]
        }
    ]
}
]

Then, the request will Continue regardless of failureMode config.

I have implemented an e2e test that reproduces this issue: #122

[didierofrivia]

I'm a bit confused, since #122 is about parallel requests to envoy and this issue is about error handling? The implementation of the operation dispatcher is a basic one, I agree we need error handling (among other things), but mistyped configuration would kind of be a thing coming from the Kuadrant Operator.
In any case, if a given request to a service fails, we should follow the failureMode config for that service and Allow/Deny respectively or define a set of operations to follow depending on the service? (i.e: retry, authorize, limit...etc)

[eguzki]

In any case, if a given request to a service fails, we should follow the failureMode config for that service and Allow/Deny respectively or define a set of operations to follow depending on the service?

I would say follow the failureMode. 🤷