How to handle errors when messages are automatically flushed? about analytics-node

[thesmart]

I have the same question. CMIIW, @dominicbarnes, but the error handling seems logically unsound. Calling track n-times, there's no way to determine which subset of track calls failed when the batch fails because all the track callbacks are called with the single batch call error. The batch method documentation contains no information about the response... does it fail all if one fails? Are payloads partially accepted?

I suppose we need to error-handle every individual call, and we won't know which specific calls failed. I suppose turning off automatic batching would help error handling but sacrifice performance.

[vadimdemedes]

I agree with @thesmart, the error handling of tracking methods (identify(), track(), etc) doesn't make much sense to me. While we could easily add flush event, I'm not sure what purpose would it serve. I'd expect analytics library to handle tracking/sending errors by itself. E.g. if flush() fails, I'd expect it to reschedule the flush and retry. But I wouldn't want to handle that logic manually.

@conorgil could you share more about your use case and what do you want to achieve with error and flush events? It would make it easier to understand the benefit of having those events.

[vadimdemedes]

Closing due to lack of response. If issue isn't resolved, feel free to reply and I'll reopen it.