HAPROXY working but on browser the connection are "sticky"

[stanthewizzard]

Hello

I'll explain the subject

Code: [Select]

#
# Automatically generated configuration.
# Do not edit this file manually.
#

global
    uid                         80
    gid                         80
    chroot                      /var/haproxy
    daemon
    stats                       socket /var/run/haproxy.socket group proxy mode 775 level admin expose-fd listeners
    nbproc                      1
    nbthread                    1
    hard-stop-after             60s
    no strict-limits
    tune.ssl.default-dh-param   2048
    spread-checks               2
    tune.bufsize                16384
    tune.lua.maxmem             0
    log                         /var/run/log local0 info
    lua-prepend-path            /tmp/haproxy/lua/?.lua

defaults
    log     global
    option redispatch 0
    timeout client 30s
    timeout connect 10s
    timeout server 30s
    retries 3
    default-server init-addr last,libc

# autogenerated entries for ACLs


# autogenerated entries for config in backends/frontends

# autogenerated entries for stats




# Frontend: FRONTEND ()
frontend FRONTEND
    bind 127.0.0.1:4443 name 127.0.0.1:4443
    mode tcp
    default_backend NPM_backend
    # tuning options
    timeout client 30s

    # logging options
    # ACL: exchange_condition
    acl acl_63185f07e1e485.01440772 req.ssl_sni -i mail.xxxx.com
    # ACL: rds_condition
    acl acl_6318646aec4140.35738389 req.ssl_sni -i rds2022.xxxx.com
    # ACL: exchange_autodiscover
    acl acl_63186489c20228.86220338 req.ssl_sni -i autodiscover.xxxx.com

    # ACTION: Exchange_rule
    use_backend Exchange_backend if acl_63185f07e1e485.01440772
    # ACTION: RDS_rule
    use_backend RDS_backend if acl_6318646aec4140.35738389
    # ACTION: Exchange_autodiscover_rule
    use_backend Exchange_backend if acl_63186489c20228.86220338

# Backend: Exchange_backend ()
backend Exchange_backend
    # health checking is DISABLED
    mode tcp
    balance source

    # tuning options
    timeout connect 10s
    timeout server 30s
    server Exchange aaaaaaaaaaaaaaa8:443 resolve-prefer ipv4

# Backend: RDS_backend ()
backend RDS_backend
    # health checking is DISABLED
    mode tcp
    balance source

    # tuning options
    timeout connect 10s
    timeout server 30s
    server RDS bbbbbbbbbbbbbbbbbbb:443 resolve-prefer ipv4

# Backend: NPM_backend ()
backend NPM_backend
    # health checking is DISABLED
    mode tcp
    balance source

    # tuning options
    timeout connect 10s
    timeout server 30s
    server NPM_PRIME ccccccccccccccccccccc:443 resolve-prefer ipv4

listen local_statistics
    bind            127.0.0.1:8822
    mode            http
    stats uri       /haproxy?stats
    stats realm     HAProxy\ statistics
    stats admin     if TRUE

# statistics are DISABLED
It works "finish"

When I connect to exchange with mail.xxxx.com
I have exchange

If I connect right after to rds2022.xxxx.com
the address is rds2022.xxxx.com/owa/etccccccccc

If I wait a bit a recharge the page with rds2022.xxxx.com
I have the RDS server

Same thing for backend

Why are the "session" (i presume) sticked for a certains time to the previous server called by URI ?

Thanks for help (hope I'm clear)

[sorano]

Question: Are you using the same wildcard certificate for both frontends?

I encountered something similar a couple of years ago and it was due to connection coalescing.

You can read more about it here:
https://daniel.haxx.se/blog/2016/08/18/http2-connection-coalescing/

If my memory serves me right I solved it by using a separate non-wildcard cert for one of the frontends.

[stanthewizzard]

There is only 1 frontend

This is the same wildcard xxxx.com for exchange, rdc and npm (backend)



I’ll read this
Thank you

[sorano]

Yes
This is the same wildcard xxxx.com for exchange, rdc and npm

I’ll read this
Thank you

Ok, then connection coalescing is the root cause of your issues for sure.

[stanthewizzard]

There is only 1 frontend
3 backend

[Fright]

Hi
is browser impact excluded?

[stanthewizzard]

it only affect browser.
if I kill safari on iOS. It's good. In fact I have to kill safari to browser between owa and nginx