haproxy victoria metrics active tcp check handshake loadbalancer

haproxy victoriametrics active tcp check handshake

If you are using Victoria Metrics with load balancer like haproxy with active healt checks, your vmstorage logs will be flooded 
with messages like this:

{"ts":"2025-07-24T10:24:00.291Z","level":"error","caller":"VictoriaMetrics/lib/vmselectapi/server.go:203","msg":"cannot perform vmselect handshake with client \"127.0.0.1:33942\": cannot read hello: cannot read message with size 11: read tcp4 127.0.0.1:8401->127.0.0.1:33942: read: connection reset by peer; read only 0 bytes"}
{"ts":"2025-07-24T10:24:02.464Z","level":"error","caller":"VictoriaMetrics/lib/vmselectapi/server.go:203","msg":"cannot perform vmselect handshake with client \"127.0.0.1:33960\": cannot read hello: cannot read message with size 11: read tcp4 127.0.0.1:8401->127.0.0.1:33960: read: connection reset by peer; read only 0 bytes"}
{"ts":"2025-07-24T10:24:04.603Z","level":"error","caller":"VictoriaMetrics/lib/vmselectapi/server.go:203","msg":"cannot perform vmselect handshake with client \"127.0.0.1:33980\": cannot read hello: cannot read message with size 11: read tcp4 127.0.0.1:8401->127.0.0.1:33980: read: connection reset by peer; read only 0 bytes"}
{"ts":"2025-07-24T10:24:06.588Z","level":"error","caller":"VictoriaMetrics/lib/vmselectapi/server.go:203","msg":"cannot perform vmselect handshake with client \"127.0.0.1:33994\": cannot read hello: cannot read message with size 11: read tcp4 127.0.0.1:8401->127.0.0.1:33994: read: connection reset by peer; read only 0 bytes"}
{"ts":"2025-07-24T10:24:08.629Z","level":"error","caller":"VictoriaMetrics/lib/vmselectapi/server.go:203","msg":"cannot perform vmselect handshake with client \"127.0.0.1:58416\": cannot read hello: cannot read message with size 11: read tcp4 127.0.0.1:8401->127.0.0.1:58416: read: connection reset by peer; read only 0 bytes"}

this is because handshake is incomplete.

Handshake is: 'vmselect.01\nok'
38402 (= vmstorage port)

# echo -ne 'vmselect.01\nok' | nc -w 1 localhost 38402



Add this to haproxy backend section:

mode tcp
option tcp-check
tcp-check send vmselect.01\nok
tcp-check expect string ok
  
and your vmstorage logs will be nice and clean