using containerized agents

[Daryl Brown]

So I have an upcoming project that will involve running some automation on a containerized pod in Azure.

My understanding of the architecture is that there'll be several pods installed on this container, and that they will be dynamically spinning up new pods and breaking down others based on demand at the time.

Now I see in the documentation about how to go about installing an agent on a container -- I haven't tried it yet, but I see the documentation there for it -- and I understand that you install your agent on the container itself, not on the pod.  But I'm not seeing anything that talks about how I would then use that agent to run a job against one of these pods.

So for example, let's say this container has five (linux) pods running on it at the moment, and I need to run a job against one of those pods.  (In the basic example I've been given, I would need to copy a file into a local input folder on one of these pods.  Each pod has it's own local input folder; they aren't using a shared folder for this.  All the pods would be essentially identical, so it wouldn't really matter which pod I use...although I might need for a downstream job in a workflow to be able to target the same pod chosen for this initial file copy.)   How do I build a job to do that, exactly?  Am I building a regular unix job (JOBS) for this, or is there a special type of container-JOBS you use for containerized agents?

I would think I would need a special type of JOBS here that would allow me to somehow specify which pod I need to use, but I haven't seen anything along those lines in the documentation.

Am I thinking about this all wrong?  I'm new to the world of containers here, so please correct me if I'm approaching this incorrectly or asking the wrong kind of questions here.

Thanks in advance!

[Markus Embacher]

Hi Daryl,

when tasks run on an agent-group (HOSTG) then the workflow (JOBP) will make sure that the agent-group will always resolve to the same agent, it the option "Assign same agent when tasks run in the same agent group" is selected. You can find that option in the "attributes" tab of the workflow.

So in your example you could group your agents in agent-groups. The assignment to those agent groups can be automated by using wildcards based on agent-name or roles (to be defined in the agents .ini file).

Regards, Markus

[Oana Botez]

Hi Daryl,

In order to run agent containers as pods, you need to create Kubernetes deployments. You can find some examples in the documentation.

BR,
Oana

[Daryl Brown]

Hi Oana,

I think this is probably what I was looking for...I just didn't recognize it as such due to my unfamiliarity with containers/pods/Kubernetes.

What's still not clear to me here: when I install a containerized agent here -- and let's say this container currently contains 5 pods on it -- would I end up seeing 5 agents in my environment (which I could then put into an agentgroup), or just one?  If just one, then when I go to execute a job on that containerized agent, does it just randomly select one of the pods in that container to run on?  If it chooses it randomly, then is there a way to target that same pod in a downstream job in a workflow?

If there is a video or some screenshots illustrating how this works, that would be extremely helpful.

[Marcin Uracz]

Hi Daryl,

In Kubernetes, pods consist of  one or more containers, not the other way around. If you have a multi-container pod the containers inside share resources like network or storage allowing for tight integration.

I think the main question is what are you trying to achieve and what is the application architecture you are trying to automate. For running Kubernetes Jobs you don't need a OS Agent there is a dedicated Kubernetes Integration (Agent)  on the Automic Marketplace. 

------------------------------
🧙 PEM Creator

Online Automic Training 24/7:
https://membership.philippelmer.com/ (deutsch)
https://pemautomic.com (english)
------------------------------