zabbix HA behind reverse proxy

[db100]

Dear all,

in a situation where 2 zabbix servers are in HA modus and reachable to the clients via a reverse proxy (haproxy / traefik, etc), is it possible for the reverse proxy to figure out which one node is the current master, so as to remove the other(s) from the load balancing routine?

example in traefik one could use healthchecks: https://doc.traefik.io/traefik/routi.../#health-check

this is quite an important feature to have

[Steven09]

Traefik supports health checks to monitor the status of backend servers. You can configure Traefik to periodically check the health of both Zabbix servers, typically using HTTP/HTTPS health checks to a specific endpoint or port exposed by the Zabbix servers.

[db100]

Steven09 thanks for the reply. as highlighted in my comment above, health-check only works with HTTP endpoints, but the zabbix-server does not expose any. i wonder whether i should use haproxy instead, but then again, would the standard tcp port return an "error" when the server is in standby mode ? or how to figure out that a server is in standby ?

[cyber]

You can find which host is active from DB, table ha_node.

[db100]

but reverse proxies cannot read stuff from DBs, thats why i am trying to figure that out via TCP protocol, is there any way to do that ?

[cyber]

If 10051 port answers, then that one is active... Inactive node does not listen on that port...

[db100]

i see, ok i will check this then, if it is possible to achieve something with traefik

[db100]

using haproxy as reverse proxy in front of 2 (or more) zabbix servers does the job. here i share the config i am using, hopefully this will benefit someone out there. Btw, you could think to add this to the official helm chart of zabbix:

# taken from: https://www.securityandit.com/network/haproxy-for-service-discovery-in-kubernetes/
global
    log stdout format raw local0 notice
    daemon
    maxconn 10000
resolvers kubernetes
    nameserver dns1 <k8s_dns_ip>:53
    resolve_retries       3
    timeout retry         1s
    hold valid            5s
    accepted_payload_size 8192
# Add stats page configuration for HAProxy GUI
listen stats
    bind *:8404  # Bind stats page to port 8404
    stats enable
    mode http
    stats uri /stats
    stats refresh 10s
defaults
    log global
    mode tcp
    timeout connect 5s
    timeout client  10s
    timeout server  10s
frontend zabbix-frontend
    bind *:10051
    default_backend zabbix-backend
backend zabbix-backend
    balance roundrobin
    option tcp-check
    timeout connect 3s
    timeout server  5s
    timeout check   5s
    server-template zabbix 3 zabbix-zabbix-server-ha.<namespace>.svc.cluster.local:10051 resolvers kubernetes resolve-prefer ipv4 check