Cluster failure scenario - containers are being migrated to another host but all the stack services crash

Rancher Version: v1.1.3
OS and where are the hosts located? VMWare RHEL 7.2 nodes, internal DB.
Environment Type: Cattle

Steps to Reproduce:

I created two nodes in my Rancher/Cattle environment.
Figure 1: s13.postimg.org/xmupg94fb/docker1.jpg

After that, I created a layer called http and underneath it a service “httpd” and loadbalancer “lb-httpd”.
The service was created with a health check “TCP Connection Opens”, on port 80 and other fields with default values.
Figure 2: s13.postimg.org/ciw4anls7/docker2.jpg

When I scale up the httpd service, the balance between the hosts happens as expected.
Figure 3: s12.postimg.org/k3hwtf225/docker3.jpg

But if I turn off the second host (systemctl stop docker on pxl0dockr006), the containers that were running on that host
(pxl0dockr006) are migrated to the first host (pcl0dockr005), but do not start properly.
Moreover, the containers that were working correctly on the first host (pxl0dockr005), crash.
Figure 4: s9.postimg.org/tmq529x6n/docker4.jpg

When I turn on the second host (pcl0dockr006), the balance between the hosts changes completely.
Figure 5: s22.postimg.org/6629mzk69/docker5.jpg

Could you help me identify the cause of the problem?

We managed to solve. In the “Add Host” the CATTLE_AGENT_IP parameter
(field number 4 of the form) must be completed for the first host (himself). For other hosts leave the field blank.