GutoVeronezi commented on PR #8605:
URL: https://github.com/apache/cloudstack/pull/8605#issuecomment-2023718630

   @andrijapanicsb, I made some changes in the PR's title and description to 
make the proposal clearer. 
   
   To put us on the same page, the proposal with this PR is to flexibilize the 
timeout of the commands MS sends to the agents. It was never aimed to provide 
API calls for operators/users.
   
   Currently, almost every command MS sends to the agents uses the global 
setting `wait` as the timeout's value. That means, if you want to extend or 
shorten the execution of a command, you will affect every other command. 
Therefore, currently we **do not** have a way to customize the timeout for 
specific commands.
   
   For some edge cases, there are specific global settings, like 
`kvm.storage.online.migration.wait` and `kvm.storage.offline.migration.wait`. 
Externalizing global settings for each command to provide flexibility is not 
interesting, as we would increase the number of global settings by almost 
50%(we would also have the work to create and describe every setting, which you 
mentioned "doesn't sound feasible"). Furthermore, if we continue working with 
global settings, each command timeout externalization would require the process 
of (i) providing the use case; (ii) extending the platform; and (iii) waiting 
for a release. We know that process can take several months and, if you are an 
operator, you know that waiting months for a simple timeout configuration is 
not feasible.
   
   The proposal is to create a mechanism to allow operators to define the 
timeout of the commands MS sends to the agents dynamically and in a flexible 
fashion. The first phase is focused on the definition of the concept; thus, it 
is expected to be more developer-friendly at first. However, we already have 
the next steps planned, as you can see in 
https://github.com/apache/cloudstack/issues/8506#issuecomment-1924648687, and 
those will make the feature user/operator-friendly (via API and GUI).
   
   Indeed, the process of documenting all the commands will require some work, 
but we have contributors who are willing to do that (me and some others); thus, 
as it will not require efforts from the operators to document them, I do not 
see the problem.
   
   It is also important to mention that it is a new feature and the current 
behavior is kept; thus, you do not need to use the feature if you do not want.
   
   If you have another effortless and more straightforward proposal that 
addresses all of the use cases we are talking about, I would be glad to discuss 
it with you.
   
   Regarding the timeout of the API calls the users do the MS, as they are a 
different context, we can create another issue to discuss it.


-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]

Reply via email to