How much should I be worried about termination status "cD" in TCP sessions

Help!
1

Hello,

outline of my problem: I see requests that are closed with termination_status cD. The number of requests is tied to the value set for configuration item timeout client (I’ve started with 2000 and getting a high number), setting timeout client 60000 decreased the frequency but seems to be high to me. The HTTP client library is python requests.

haproxy version is 1.8.8

question:

  • How much should I be worried about such status considering I got around 10 errors for about 17800 requests per hour.
  • Is my configuration wrong ?
  • do you have tips to debug such issue ?

here my setup

  • haproxy is installed locally on each node (7) that requires external access
  • a frontend fe is listening in tcp mode and send request to backend be
  • a backend be is configured to have 2 backend servers which are squid
  • application on each node is configured to use http://127.0.0.1:18888 as proxy

overview.

┌───────────────────────────┐                     ┌───────────┐
│  ┌─────────────────────┐  │                     │           │
│  │                     │  │       ┌─────────────►  squid-1  ├─────────►
│  │  local application  │  │       │             │           │
│  │                     │  │       │             └───────────┘
│  └─────────┬───────────┘  │       │
│            │              │       │             ┌───────────┐
│  ┌─────────▼───────────┐  │       │             │           │
│  │    haproxy          │  │       │             │  squid-2  ├──────────►
│  │  (localhost:18888)  ├──┼───────┴─────────────►           │
│  └─────────────────────┘  |                     └───────────┘
└───────────────────────────┘

config excerpt:

defaults
    log global
    maxconn 40000
    mode tcp
    option dontlognull
    option redispatch
    option tcplog
    option log-health-checks
    retries 3
    timeout client 60000
    timeout connect 10000
    timeout queue 30000
    timeout server 90000
    timeout check 2000
    default-server inter 1s fall 3 rise 2
...
frontend fe
    mode tcp
    bind 127.0.0.1:18888
    default_backend be
...
backend be
    mode tcp
    balance roundrobin
    option tcplog
    option tcp-check
    tcp-check connect
    server squid-1 squid-1:8888 check
    server squid-2 squid-2:8888 check

Thanks