You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Currently Aragorn implements async callbacks using celery queues.
Once a async request is sent off to strider , a background worker is issued to listen to a celery queue. When strider calls back it targets that particular queue, and message is transferred using that queue.
We can think of scaling this up with multiple striders and multiple Aragon instances. When thinking in that direction the problem becomes that Aragon replicas would need a shared queue, and even then queue id management becomes a shared concern of all Aragorn , to keep track of which queue belongs to which instance.
Here we can propose having a smaller paired solution where Aragorn and strider are deployed as separate containers in a single pod. This would mean localhost routing between them is possible. And they would be able to communicate by just swapping ports and using localhost. This will guarantee that strider calls the Aragorn instance it initiated it , and vice versa.
As a start we could think about restructuring this strictly from deployment perspective, i.e have the aragorn pods a celery and a strider container aswell. Once that is working
Then replace celery solution to a more light weight solution now that we have everything basically locally with in a single pod. in the code.
The text was updated successfully, but these errors were encountered:
Currently Aragorn implements async callbacks using celery queues.
Once a async request is sent off to strider , a background worker is issued to listen to a celery queue. When strider calls back it targets that particular queue, and message is transferred using that queue.
We can think of scaling this up with multiple striders and multiple Aragon instances. When thinking in that direction the problem becomes that Aragon replicas would need a shared queue, and even then queue id management becomes a shared concern of all Aragorn , to keep track of which queue belongs to which instance.
Here we can propose having a smaller paired solution where Aragorn and strider are deployed as separate containers in a single pod. This would mean localhost routing between them is possible. And they would be able to communicate by just swapping ports and using localhost. This will guarantee that strider calls the Aragorn instance it initiated it , and vice versa.
The text was updated successfully, but these errors were encountered: