Add ank apply command to CLI

[lingnoi]

Issues: #26

Added ank apply command

Definition of Done

The PR shall be merged only if all items mentioned in CONTRIBUTING.md have been followed. In case an item is not applicable as described, please provide a short explanation in the description.

• All steps in CONTRIBUTING.md have been handled

[inf17101]

Maybe one suggestion and meant optionally (It expects how much workloads we expect on average touched with this new command): The ank apply command outputs the workloads it has touched to the console within { ... } which is nice. However, if there are many workloads maybe the output is too much in the console and not helpful anymore. Maybe we can limit the output to a certain number and output then { w1, w2, w3, ... } or showing only in verbose mode all of them if someone needs that really for debugging.

[windsource]

Maybe one suggestion and meant optionally (It expects how much workloads we expect on average touched with this new command): The ank apply command outputs the workloads it has touched to the console within { ... } which is nice. However, if there are many workloads maybe the output is too much in the console and not helpful anymore. Maybe we can limit the output to a certain number and output then { w1, w2, w3, ... } or showing only in verbose mode all of them if someone needs that really for debugging.

@inf17101 Not sure what you mean. When I call ank apply it only returns the names of the affected agents and not of the workloads.

❯ ank apply databroker.yaml
agent_names={"agent_A"}
error: ExecutionError("Failed get response from server in time (timeout=3s).")

However, I do not think that we shoud limit those.

[inf17101]

Maybe one suggestion and meant optionally (It expects how much workloads we expect on average touched with this new command): The ank apply command outputs the workloads it has touched to the console within { ... } which is nice. However, if there are many workloads maybe the output is too much in the console and not helpful anymore. Maybe we can limit the output to a certain number and output then { w1, w2, w3, ... } or showing only in verbose mode all of them if someone needs that really for debugging.

@inf17101 Not sure what you mean. When I call ank apply it only returns the names of the affected agents and not of the workloads.

❯ ank apply databroker.yaml
agent_names={"agent_A"}
error: ExecutionError("Failed get response from server in time (timeout=3s).")

However, I do not think that we shoud limit those.

If you would not have the response errors (because of missing @christoph-hamm' PR) then it would print the workloads as well and potentially spam the output if there are many to touch. But it is anyway meant optionally. But I thought it is maybe helpful to mention, because today's applications would handle such case. I think it is not a bad idea to restrict output that can be not calculated how much it would be.

[windsource]

Supporting multiple agents in yaml files provided to ank apply is required.

[lingnoi]

It is not possible to deploy workloads on different agents.

❯ ank apply *.yaml
error: Multiple agent names in manifests detected ["agent_A", "infotainment"] -> use '--agent' option to overwrite! 

I do not understand that message. Why should it be forbidden to deploy workloads on different agents and how does the option --agent help here?

Currently it is not supported and you need to tell via the --agent option on which agent to apply the workloads. If it is required then I will adapt the code to support multiple agents provided in the workload specs. Is it required?

[windsource]

It is not possible to deploy workloads on different agents.

❯ ank apply *.yaml
error: Multiple agent names in manifests detected ["agent_A", "infotainment"] -> use '--agent' option to overwrite! 

I do not understand that message. Why should it be forbidden to deploy workloads on different agents and how does the option --agent help here?

Currently it is not supported and you need to tell via the --agent option on which agent to apply the workloads. If it is required then I will adapt the code to support multiple agents provided in the workload specs. Is it required?

When an application developer provides an application with workloads for different agents then one manifest should be enough to describe the deployment of that application. So I think it required.

[windsource]

Currently the output does not look very clear. What about using a table structure as output?

$ ank apply file1.yaml file2.yaml
WORKLOAD    AGENT    STATUS  FILENAME    
databroker  agent_a  OK      file1.yaml  
nginx       agent_b  Error   file1.yaml  
db          agent_b  OK      file2.yaml  

[lingnoi]

Currently the output does not look very clear. What about using a table structure as output?

$ ank apply file1.yaml file2.yaml
WORKLOAD    AGENT    STATUS  FILENAME    
databroker  agent_a  OK      file1.yaml  
nginx       agent_b  Error   file1.yaml  
db          agent_b  OK      file2.yaml  

Okay, I like your suggestion! I understood that the current output contains some kind of redundant information. I will adapt it.

[maturar]

I found a bug when I have tried to overwrite the agent with the --agent flag.
I have created workloads with the command ./ank apply --agent agent_B manifest.yaml. It means that the flag --agent agent_B ovewrites the configuration in the manifest.yaml (here is the agent set to agent_A. The command produces following ouptut:

 WORKLOAD NAME   AGENT     OPERATION    STATUS   FILE          
 hello1          agent_A   Add/Update   OK       manifest.yaml 
 nginx           agent_A   Add/Update   OK       manifest.yaml 

But in the reality the Ankaios creates workloads in agent_B:

$ ./ank get workloads
 WORKLOAD NAME   AGENT     RUNTIME   EXECUTION STATE   ADDITIONAL INFO 
 hello1          agent_B   podman    Failed(Lost)                      
 nginx           agent_B   podman    Running(Ok)                       

The information printed by the ank apply is not consistent with the real state.

The output of ank apply .. is not the same as ank get workloads. As it is a diffrent command. The "STATUS" of ank apply .. shall not represent the "EXECUTION STATE", instead it represent the apply status. Why should this status be the same as execution state?

Well for me as a user it is confusing if I say that I would like to overwrite the agent name with the flag and the command answers with something else. My understanding is that the table shall represent a summary what information the command ank apply is going to use.
BTW: I was reading all comments here and I found out this comment: #180 (comment) Is not this the same issue?

If the agent name is meant then it is the same issue!

The issue has been fixed.

[windsource]

I have created two workloads with:

❯ ank apply --agent infotainment speed-consumer.yaml speed-provider.yaml
 WORKLOAD NAME    AGENT          OPERATION    STATUS   FILE
 speed-consumer   infotainment   Add/Update   OK       speed-consumer.yaml
 speed-provider   infotainment   Add/Update   OK       speed-provider.yaml

and the output looks OK. But when I delete those workloads with:

❯ ank apply -d speed-consumer.yaml speed-provider.yaml
 WORKLOAD NAME    AGENT     OPERATION   STATUS   FILE
 speed-consumer             Remove      OK       speed-consumer.yaml
 speed-provider   agent_A   Remove      OK       speed-provider.yaml

the agent names are wrong. As the agent names are not known they column AGENT shall be removed for the output of ank apply -d .

[windsource]

OK from a user perspective. Still some approval from technical perspective required.

[maturar]

There is a merge conflict in this PR. @lingnoi could you resolve it, please?

[maturar]

Okay for me. @christoph-hamm is reviewing fixes of his findings in the swdesign.

[christoph-hamm]

Only two small findings.