Stopping a running workload or triggering its failed condition

[mrogonna]

Hey guys,
My observation so far(correct me if i'm wrong):

1. you can only add a new workload or delete any workload over CLI, no option to temporary stop/start or trigger failed condition for the workload
2. if you start an agent, it automatically starts all the workloads under that particular agent unless there are some inter-workload dependencies that needs to be met

Question:

1. how can I trigger a failed condition (exit with non-zero return code) for a workload to fulfil one of the inter-workload dependencies, ExecutionState(Failed(ExecFailed)) => AddCondition(ADD_COND_FAILED)
2. for debugging purposes, how can I manually stop/start a single running workload over cli?
3. is there a reset key? if yes, how can I reset the program to start from my initial startup configurations?

thank you for the help.

[windsource]

Hi @mrogonna,

the following Ankaios manifest shows how to trigger a workload based on failed workload:

apiVersion: v0.1
workloads:
  willfail:
    runtime: podman
    agent: agent_A
    runtimeConfig: |
      image: docker.io/busybox:latest
      commandArgs: ["sh", "-c", "exit 1"]
  onfail:
    runtime: podman
    agent: agent_A
    dependencies:
      willfail: ADD_COND_FAILED
    runtimeConfig: |
      image: docker.io/busybox:latest
      commandArgs: ["sh", "-c", "echo 'I have been triggered.'"]

Save this to a file example.yaml and then call ank apply example.yaml.

For debugging purpose you can use ank apply FILENAME to start workloads and ank apply -d FILENAME to delete workloads.

The startup configuration is only applied at startup. So if you have a running Ankaios cluster and want to start a clean setup, perform the following steps:

1. Delete all running workloads using ank delete <list of all workloads>
2. Stop Ankaios agent by sudo systemctl stop ank-agent
3. Stop Ankaios server by sudo systemctl stop ank-server
4. Change the startup config
5. Start server and agent again

[mrogonna]

thank you for the promt reply @windsource
I understand the manifest example.yaml above fails immediately after its started because of the runtime config:

runtimeConfig: |
      image: docker.io/busybox:latest
      commandArgs: ["sh", "-c", "exit 1"]

but my question is how to achieve the failed condition over CLI so its easy to stop and restart any workload without deleting it, this is useful for other applications/workload I intend to run for some time before failing..

My Idea if possible is that Ankaios would be useful to implement High Availability Cluster were a particular workload will be available to take over a failed one

please let me know if you need more clarity on my question, I really appreciate.

[windsource]

If the workload should be restarted when it fails, Ankaios defines the restartPolicy: ALWAYS settings. Note that it is not implemented yet but will be in the next release.

In case you want to interact with a running workload you need to use the runtime tooling. So for the runtime podman you might use podman container stop to stop a workload. It is up to the workload to return the exit code when it received the signal to terminate.