Rancher’s core scheduling logic is part of Rancher, which handles port conflicts and ability to schedule based on labels on hosts/containers. In addition to the core scheduling logic, Rancher supports additional scheduling policies by using the Rancher Scheduler catalog item.
Note: These functionalities are not available for Kubernetes as Kubernetes handles scheduling of their own pods.
Rancher scheduler is defaulted to be
required for any environment template. If you have deleted this stack from your environment, you can add it back into the environment by launching Rancher Scheduler in Catalog -> Library.
By default, Rancher assumes that there is only one IP on the host that is available when scheduling services that publishes ports or launching load balancers. If your host has multiple IPs that could be used, the host needs to be configured to allow the Rancher scheduler to know which IPs are available.
When there are multiple IPs available on the host for scheduling, Rancher will schedule against all the available scheduler IPs when a port is published through a service or a load balancer. The scheduler will report port conflicts after all available scheduler IPs on the host have been allocated for that port.
When Rancher hosts are added into Rancher, they are automatically assigned resource limits based on the size of the host. These limits can be adjusted by editing the host. In Infrastructure -> Hosts, you select Edit from the dropdown of the host. In the Resource Limits section of the host, you can update the Memory or CPU for the maximum amount that you want to be used on the host.
When creating a service, you can specify the Memory Reservation and mCPU Reservation in the Security/Host tab. When these reservations are set, containers of the service are only scheduled onto hosts that have available resources. The maximum limit of these resources on a host are determined based on the Resource Limits of the host. If scheduling the container onto the host would force these limits to be over the threshold, then the container will not be scheduled onto the host.
version: '2' services: test: image: ubuntu:14.04.3 stdin_open: true tty: true # Set the memory reservation of the container mem_reservation: 104857600
version: '2' services: test: # Set the CPU reservation of the container milli_cpu_reservation: 10 scale: 1
Typically, most of the container scheduling is defined on the service. The service will have specific rules or restrictions on which host that the container(s) can be scheduled on. For example, a container must be scheduled onto a host that has a specific host label. Rancher has the ability to specify requirements on the host to only allow specific containers to be scheduled onto the host. One example of when you might want to use this feature is if you wanted to dedicate a host only have database containers scheduled onto the host.
Note: When you add label restrictions for containers on a host, you will be required to include a label in order for our infrastructure services to be scheduled onto the host. Without these services, containers required to allow networking and other key components of Rancher to work will not be scheduled onto the host.
For any host, you will edit the host to add in what labels a container must have by selecting Edit from the dropdown of the host. In the Require Container Label section, you can add which labels are required to be on a service in order for those containers to be scheduled onto the host. The UI will automatically have the label (i.e.
io.rancher.container.system=) as a required label.