I managed to automate (with ansible) the installation of Rancher HA with 2 servers behind nginx, 1 db and 3 rancher-agent hosts. It works great, the lb does it’s job, the db as well. But I see a strange problem that was not happening when having a non-HA setup (1 server, 3 rancher-agent hosts). The infrastructure stacks take about 5 to 10 minutes to become active. I though it was the LB that had weird behaviours so I explicitly create the rancher-agent hosts on one of my 2 server setting the CATTLE_URL env var. I have the exact same behaviour.