CouchDB Haproxy With Quick Client Disconnect

haproxyabrt.cfg

global
    maxconn 10000
    spread-checks 5

defaults
    mode http
    log global
    option httplog
    log stdout format raw local0
    balance roundrobin
    option redispatch
    retries 4
    timeout client 150s
    timeout server 360s
    timeout connect 500

    option   abortonclose

frontend http-in
    option clitcpka
    clitcpka-cnt   3
    clitcpka-intvl 1s
    clitcpka-idle  3s

    bind *:5984
    default_backend servers

backend servers
    option httpchk GET /_up
    server node1 127.0.0.1:15984 check inter 600s

CouchDB logs show the request process is killed within 5 seconds or so

[notice] 2023-09-27T03:03:22.837658Z node1@127.0.0.1 <0.18737.0> 47b8291a4c 127.0.0.1:5984 127.0.0.1 adm PUT /db?q=1 201 ok 114
[notice] 2023-09-27T03:03:22.920781Z node1@127.0.0.1 <0.18737.0> e1ca592e21 127.0.0.1:5984 127.0.0.1 adm POST /db/_bulk_docs 201 ok 79
[notice] 2023-09-27T03:03:22.999434Z node1@127.0.0.1 <0.18737.0> fdbba7a0ee 127.0.0.1:5984 127.0.0.1 adm POST /db/_bulk_docs 201 ok 75
[notice] 2023-09-27T03:03:23.075838Z node1@127.0.0.1 <0.18737.0> f9af5e4d9a 127.0.0.1:5984 127.0.0.1 adm POST /db/_bulk_docs 201 ok 73
[notice] 2023-09-27T03:03:23.152235Z node1@127.0.0.1 <0.18737.0> 888ea8db06 127.0.0.1:5984 127.0.0.1 adm POST /db/_bulk_docs 201 ok 73
[error] 2023-09-27T03:03:23.155898Z node1@127.0.0.1 <0.18737.0> b311b36392  ++++++++++TRACING _find++++++++ mango_httpd:handle_find_req@208 REQPID:<0.18737.0>
(<0.18737.0>) spawn <0.19452.0> as erlang:apply(#Fun<rexi_monitor.1.77517421>,[])
(<0.18737.0>) link <0.19452.0>
(<0.18737.0>) spawn <0.19453.0> as erlang:apply(#Fun<fabric_streams.3.91837225>,[])
(<0.18737.0>) exit {shutdown,client_disconnected}

Capture traffic from both client <-> haproxy:5984 and haproxy <-> couchdb:15984 connections

tcp port 5984 or tcp port 15984