Cant scale up service

Rancher 1.x
1

Updated rancher to the 0.51.0 from 0.47.0 with the UI changes, I cant seem to be able to scale up my services. Previously in the old UI, you could just hit the scale up button. Is this still possible via rancher?

2

,

Pasted image1058×395 20 KB

or

Pasted image1560×171 19.7 KB

3

Right, looks like the way scaling works has changed. I had my web container scheduled to global and I previously was able to scale up as I pleased. That does not work anymore. This is my configuration below.

`redis:
labels:
io.rancher.scheduler.affinity: web
image: redis:2.8
privileged: true
cron-web:
ports:
- 443/tcp
- 80/tcp
- 8080/tcp
environment:
SOME_ENV_DETAILS: value
labels:
io.rancher.scheduler.affinity:container_label: co.seeddigital.type=web
io.rancher.container.pull_image: always
command:
- /run_cron.sh
image: seed/myapp_web:latest
links:
- redis:myredis
privileged: true
web:
ports:

  • 443/tcp
  • 80/tcp
  • 8080/tcp
    environment:
    SOME_ENV_DETAILS: value
    labels:
    io.rancher.scheduler.global: 'true’
    io.rancher.container.pull_image: always
    io.rancher.scheduler.affinity:host_label: env=production
    co.seeddigital.type: web
    image: seed/myapp_web:latest
    links:
    • redis:myredis
      privileged: true
      lb:
      ports:
    • ‘80’
    • 443:8080
      labels:
      io.rancher.scheduler.global: 'true’
      io.rancher.loadbalancer.ssl.ports: '443’
      image: rancher/load-balancer-service
      links:
    • web:web`

Now I don’t get the ± signs to scale, all I get is Scale:Global
@vincent any help would be much appreciated. Cheers,

4

Maybe there was some build where they were incorrectly shown or something, but this has never been intentionally true. Global means “run one container on each host that matches the scheduling labels”. If you want more than one per host then you want regular scale with a soft anti-affinity rule to itself (“should not have a container with the label ${stack_name}/${service_name}”).

5

Sorry mate, I don’t get what I am supposed to do here. Change the scheduling rule ?

6

Yes, you’ll need to change your scheduling rule.

A global scale is meant to launch 1 container on every host. If you add a scheduling rule, we anticipate it to be a hard scheduling rule. There is no concept of using the global scheduling rule to launch containers onto a host that might not have your host label.

If you are looking to spread out your containers in a service across hosts, you’ll have to specify your scale and then use the label that @vincent suggested.

web:
  labels:
    io.rancher.scheduler.affinity:container_label_soft_ne: io.rancher.stack_service.name=$${stack_name}/$${service_name}
7

Awesome thanks for the info @denise turns out the reason it wasn’t working was we where using ports instead of expose for our services. Using ports meant that web containers where trying to bind to the same port causing conflict.

Thanks guys