We noticed an unexpected (at least for us) behavior of newrelic’s python agent: trying to record custom events on a long task fails (in our example a flask request that is running ~4min before response). We record the custom event in the end.
Debugging it, the only difference I was able to spot with the short request (where recording works) is that retrieving the current transaction fails in the long-running case but works for short-running case. Therefore, the custom event is dropped (no application provided).
Is there some sort of TTL/timeout parameter for the transaction cache?
I’m not sure of what’s the best way to move on from that, but those long requests are especially interesting to monitor, and it was hard to even spot that some were dropped.
Any idea welcome !