Hi all.
I have a set of server, that are set offline as necessary with the
"disable-on-404" rule.
Some servers are put offline just before performing a server restart,
some others should remain offline before and after the restart.
Using the labels of the web statistcs page, this is a state transition
that almost always works
L7OK/200 -> L7OK4C/404 -> L4CON -> L7OK/200
The next transition, where the server should remain offline after the
restart too, is the most problematic one and almost never happens
L7OK4C/404 -> L4CON -> L7OK4C/404
In this case the transition very often is
L7OK4C/404 -> L4CON -> L7STSC/404 (red line in the web interface)
If I reset haproxy, the status is again the expected one L7OK4C/404.
I am a bit concerned because of the... "almost always": I cannot
replicate the behaviour in a controlled way.
The system is a virtual machine and I think the clock drifts due to the
VMWare scheduling mechanism can introduce some trouble.
Is this a kind known behaviour? How can I better investigate the issue?
Where should I look for other informations about what it is happening?
Thank you.
.marcoc
Some system details:
VMWare ESX
1 vCPU
8GB RAM
Ubuntu Server 12.04.01
Haproxy 1.4.18
Part of haproxy configuration:
----
global
spread-checks 4
defaults
log global
mode http
option forceclose
option httplog
option dontlognull
retries 3
option redispatch
timeout connect 2s
timeout client 50s
timeout server 50s
timeout check 5s
default-server weight 20 inter 30s fastinter 10s downinter 60s
frontend common localhost:8081
backend XXXXXX
balance roundrobin
option httpchk GET /radware/rad.html HTTP/1.1\r\nHost:\ XX.XX.XX
http-check disable-on-404
appsession ASP.NET_SessionId len 52 timeout 30m prefix
server A1 192.168.101.1:80 check
server A2 192.168.101.2:80 check
server A3 192.168.101.3:80 check
server A4 192.168.101.4:80 check
server A5 192.168.101.5:80 check
server A6 192.168.101.6:80 check
server A7 192.168.101.7:80 check
server A8 192.168.101.8:80 check
---
