we see some discrepancies in our logs for the timings. The request itself takes about 500ms but in the haproxy logs we see higher times. This happens for http/2.0 requests and not for http/1.1. The value we logged was from %Tt.
We read in the docu that this could be the case for http/2.0 and with connection reuse and included idle times in the overall value.
So we switched to %Tu, but this value is the same as %Tt.
Then we started logging all timing values and saw that for %Ti is logged with -1. The docu says “A value of -1 indicates that nothing was received on the connection.”
We think this can’t really be as there was something sent over this connection as we got a result back.
Attached are the timing values for bot requests:
We think the time is spent idling but it is not visible in the logs.
Where can we see where this time is spent?
What value should show us the real-time for the request also in case of connection reuse?
We use haproxy 2.4.9.
Thanks in advance.