Open ronanb3 opened 4 years ago
The issue comes from the haproxy configuration.
Here is the configuration working.
backend wrkhttp
mode http
balance roundrobin
option forwardfor
server master-1 10.99.92.62:80 check
server master-2 10.99.92.63:80 check
server master-3 10.99.92.64:80 check
backend wrkhttps
mode tcp
balance roundrobin
server master-1 10.99.92.62:443 check
server master-2 10.99.92.63:443 check
server master-3 10.99.92.64:443 check
My inventory configuration
[all:children]
masters
workers
...
[masters]
192.168.1.101 host="master-1" mac="00:50:52:54:60:04"
192.168.1.102 host="master-2" mac="00:50:52:54:60:05"
192.168.1.103 host="master-3" mac="00:50:52:54:60:06"
[workers]
If I remove workers totally there is an error in one of the script.
Tested installation with just masters and this seems to work correctly. We will make the following changes:
Add new property to the inventory file to allow scheduling on the masters (in effect skipping one of the tasks in the ocp-ignition.yaml
If no workers have been defined, make generate the haproxy.cfg
referencing the masters for ports 80 and 443
Starting with OCP 4.5, a configuration with 3 nodes. I did an inventory keeping a worker list and removing any entry in it. It fails if I remove the worker entry.
The cluster builds well but I cannot make the console works. It seems there is an issue with the routing.