Instances stuck in "networking" state when rancher-server ip address was changed.

[sangeethah]

Server version - V0.35.0-rc5

Had a setup with 1 host that had instances running successfully.

Rebooted rancher-server. This resulted in rancher-server getting a new ip address in my case. Rancher server was an Ubuntu instance in GCE.

After this hosts went into "reconnecting" state.

Did host registration again to set the ip address to the current rancher-server ip address.
Got the rancher-agent command to add custom host from UI and execute this on the host.

Host got to "active" state.

But I was not able to starting any new containers on the host. Containers got stuck in "networking" state.

"CATTLE_CONFIG_URL" parameter in the Network agent was still pointing to the old ipaddress of rancher-server.

Deleted the network agent from UI. This resulted in a new Network agent being created.

"CATTLE_CONFIG_URL" parameter in the Network agent now points to the new rancher-server ip address.

Attempting to start new instances still fails. Instances are stuck in "networking" state forever.

[sangeethah]

The use case is very similar to #1958 , though there is no upgrade involved in my setup.

[sangeethah]

Tested with server version - v0.37.0-rc1

Had a setup with 2 hosts that had multiple instances running and had 1 LB agent.

Rebooted rancher-server. This resulted in rancher-server getting a new ip address in my case. Rancher server was an Ubuntu instance in GCE.

After this hosts went into "reconnecting" state.

Did host registration again to set the ip address to the current rancher-server ip address.
Got the rancher-agent command to add custom host from UI and execute this on both the hosts.

Host get to "active" state.

Able to start new containers on the host.
Able to stop existing containers on the host , which restarts on service reconcile.
LB service continues to work.
Scale up Lb service . New LB agent is also functional.