I am trying to launch a kubernetes cluster over EKS which would have multiple pods in it . Once the worker node has maximum pods on it running then a new node launches and the extra pod launches over the new node. Launching of a new node takes time and creates a downtime which I want to reduce. Pod disruption budget is one option but I am not sure how to use it with scaling up of nodes.
Asked
Active
Viewed 122 times
1
1 Answers
2
A simpler way to approach this would be to have your scaling policies pre-defined to scale up at reasonably lower limits. This way, let's say if your server reaches 60% of the capacity and triggers a scale up - you would have enough grace time to not face a downtime (since the first one can handle requests while the second one bootstraps) and allow the the new server to come up.

Anshul Verma
- 1,065
- 1
- 9
- 26
-
Okay this also looks good and probably I would try to implement the same but is there any direct configurational changes which can be done to reduce the downtime like setting a specific value which would launch the machine in 40 to 50 seconds or something like that? – Kaustubh Nov 28 '20 at 17:46
-
@Kaustubh No there's no such configuration you can provide. Hopefully you are not providing any bootstrap commands to the server under `user-script`. But it takes what it takes. – Anshul Verma Nov 28 '20 at 17:53
-
No I am not using any bootstrap commands as such. Okay, then I have this option to go with. Thankyou for the same, it was very helpful. – Kaustubh Nov 28 '20 at 20:22