Note: With our new infrastructure services, we no longer launch a network agent and have moved to a specific HAProxy image for load balancers. Therefore, we no longer need rancher/agent-instance.
Starting with version 1.2.0, Rancher no longer supports AWS ELBs and only supports AWS ALB (Application Load Balancers). Please update your high availability setups to use ALBs.
How to upgrade to 1.2
Rancher 1.2 introduces some major changes to how network is managed, most notably the refactoring on the IPsec networking to a proper CNI plugin and the introduction of a new v2 LB service to provider users with more flexibility in HAProxy configurations. Due to these changes, the upgrade process
will result in network downtime and will require you to upgrade each environment to restore connectivity. The upgrade process has been split into a Rancher update followed by individual environment updates so please follow the instructions below to properly upgrade your current version to 1.2:
Rancher Server Upgrade from v1.1.4
Make sure you back up your database. You will not be able to rollback to a previous version that has been used with Rancher 1.2.x. If you need to roll back, you will only be able to use a snapshot of a database that had been running your previous version. Once you have created a backup of your database, please proceed with the normal upgrade process as per the docs.
Note: If you are using AWS security groups, please make sure ICMP is enabled in your security group.
After your Rancher server has been upgraded, you will no longer be able to access your environments until you successfully upgrade them to the 1.2 environments. Due to the network and LB changes with 1.2, the upgrade process will require downtime until the new network components have been updated and migrated. We have provided a convenience feature for you to decide when to update these environments by displaying an Upgrade Now screen for each environment that requires an update. Until you perform the update, the containers should remain functional but no management capability will be allowed, but anything trigger recreation of containers (health checks) may cause these containers to no longer work. We highly suggest you upgrade the environments as any functionality that require Rancher involvement such as HA, DNS programming, healthchecks, etc. may not work properly until it has been completed.
Once you click “Upgrade Now”, Rancher will proceed to upgrade the environment. Please be patient as this can take up to 10-20 minutes depending on the size of your environment. The environment will be successfully updated when all the stacks found under Stacks -> Infrastructure are in the active state.
For those of you that have Kubernetes environments, you must upgrade your existing k8s v1.2.6 stack to v1.4.6 after all infrastructure services are in an
active state. Note: When upgrading k8s, please be aware that there is a known issue that may cause existing pods to get deleted and recreated. If the pod is not part of a replication controller, it will not be recreated. Please plan accordingly. Again, the upgrade process may take upwards of 5-10 minutes depending on your environment and will be completed when the stack is in an
Known Limitations with Upgrade
- Upgrading Swarm environments are not supported. Due to the change from Docker regarding moving Swarm into the Docker engine, we have updated our original Swarm support to the latest Docker 1.12 Swarm.
- There are some catalog entries that are migrating folders in order to show up for environment template options. These catalog entries do not support roll back to the old entries. Examples include but not limited to: Kubernetes, all External DNS entries.
- During the upgrade from v1 load balancers to v2 load balancers, any rules using selectors will not be upgraded. These rules would need to be added into the load balancer after the environment upgrade.
- Starting with 1.2, Rancher will no longer pull stats from cadvisor but rather from docker stats. Please be aware that this will cause existing catalog items that rely on cadvisor such as Prometheus to no longer work until they have been fixed to rely on docker stats instead.
Known Major Issues
- Individual container links are not resolvable, note this is only for container links, service links are still working as expected. [#6584]
- Self signed certs do not work with Rancher server [#6122]
- Hosts in AWS created in a prior release using the UI (aka docker-machine) are not cleaned up properly when deleted from the UI [#6750]
Major Bug Fixes since v1.2.0
Fixed an issue where boot2docker hosts are known to have issues with
rancher/plugin-manager:v0.2.12, there is a an updated network services
rancher/network-manager:v0.2.13. If you have an “Upgrade Available” button next to Network Services stack, please upgrade. [#6874]
Fixed an issue where docker doesn’t have to be installed at
var/lib/dockerin order for networking to work. [#6897]
Fixed an issue where you can configure your docker bridge from the default
docker0in either the
network-servicesinfrastructure service. This can be configured when creating environment templates to an alternative location. [#6896]
Fixed an issue where the UI would lock up [#6995]
Fixed an issue where Rancher service discovery didn’t have the ability to use the search path of the host before using the Rancher domain. [#7010]
Fixed an issue where load balancers couldn’t target external services using hostname [#2624]
Fixed an issue where you can see the logs of stopped containers [#6442]
Fixed an issue where you couldn’t see the nodes in a HA setup [#6814]
Fixed an issue where a load balancer targeting the same service was exporting with a stack name [#6829]
Fixed an issue where specifying a host IP for the load balancer was missing in the UI [#6852]
Fixed an issue where container cannot ping its own hostname [#6855]
Fixed an issue where high CPU usage was occurring due to to tables related to the old network agents [#6857]
Fixed an issue where the custom config for haproxy was being sorted [#6888]
Fixed an issue where docker images from GCR were not being able to be pulled [#6916]
Fixed an issue where haproxy configs for load balancers were displaying null after upgrade [#6921]
Fixed an issue where interpolation wasn’t occurring in the volumes key [#6936]
Fixed an issue where the host registration URL wouldn’t accept
Fixed an issue where proxy environment variables for Rancher agent were case sensitive [#7019]
Fixed an issue where the load balancer would quickly fill with logs [#7028]