Hi Patrick !
Thanks a lot for providing this info !
- When working: 4 requests before the "ready" state
- When non-working: 2 requests before the "ready" state
It will be helpfull for such two cases to dump in the log its headers
and to send these logs to us along with the configuration, if you could.
To enable the logging of headers you could add the following captures
below. This will help a lot to compare.
frontend fe_transparent from unnamed_defaults_1
...
option httplog
capture request header Host len 20
capture request header Content-Length len 10
capture request header User-agent len 20
capture request header X-Forwarded-For len 20
capture response header Server len 20
capture response header Content-Length len 10
capture response header Cache-Control len 8
capture response header Via len 20
capture response header Location len 20
When you've added this captures, you need to start haproxy with -d :
./haproxy -d -f ...
To have a full picture it will be helpful to have full logs for both
cases (work/not-work) between the described Kibana states "available"
--> "ready".
As for tcpdump, If you could obtain decrypted tcpdumps for both cases,
this will be very helpfull as well.
Could you please also dump the stats with 2.8.11 and with 2.9.0 and send it.
(echo "show stat" | socat /tmp/haproxy.sock stdio > /tmp/stats_2.8.11)
Many thanks in advance,
BR
--
Bien cordialement,
Valentine
