Open smyrman opened 3 years ago
I like the main idea of this proposal, but it makes a quite big assumption: it assumes that most source system is Pub/Sub (or similar with acknowledging mechanics) which is the case for our existing integrations using Telegraf, via Google Pub/Sub to Node-RED. But this a legacy strategy, all new integrations are implemented in Node-RED directly. And whenever existing integrations require an update, I will always consider migrating the integrations to Node-RED.
We should definitively adapt, evolve this.
There seam to be some "ack" nodes in node-red:
We could have a look at what they are doning, and which message fields they are inspecting.
Somehow related (if splitting a message!)
I think it would be OK to ignore split messages (sequences) in the initial version. That is, for a split message to be acked correctly, it must be joined before it's passed to the input node.
This is suggested as an extension to #36.
We might consider to change the node output to not contain raw requests, but an aggregate of the
msg.ack
field.Suggestion:
msg.ack
for data updates on success.msg.ack
for meta data updates on success.I.e. consider the following messages:
The aggregated output 1 on success is:
{msgs: [1,2,3], ack:["a1","a1-2","a3"]}
// all messages with data.The aggregated output 2 on success is:
{msgs:[3,4], ack: ["a3", "a4"]}
// all messages with meta data.Errors, are logged, while the detailed response is discarded both on success and on error.