Istio 101 : Circuit breaker for the pods


We have the below destination rule for routing traffic to the containerized application "application1" :


We create VirtualService so that all the traffic goes to our application "application1":


In the above VirtualService all traffic goes to "application1". The below DestinationRule has an effect on "application1" and all its traffic.

To avoid having lot of requests going to our pods and slowing down the application running inside, we set a circuit breaker that will stop requests when a pod is handling more than two requests at the same time - http1MaxPendingRequests -.

We can create a circuit breaker for our service using a destination rule as follows:


Above is the configuration file for a circuit breaker for any client applications making  a request to the "application1"  of the "svc1.ns1.svc.cluster.local" service. 

We are putting a cap on the number of connections and number of
pending requests to "2".

Below is a short explanation of some of the parameters used in the above Yaml file:
  • BaseEjectionTime : the maximum ejection duration for a host before getting taken into account again
  • ConsecutiveErrors : errors before a host is ignored from the loadbalancing pool.
  • Interval : the interval for looking into the ejected pod.
  • MaxEjectionPercent : percentage of hosts that can be ejected from the load balanced pool.

Comments

Post a Comment

Popular posts from this blog

Istio 101 : How Istio ServiceEntry relates to Kubernetes Services and Endpoints

Image
Kubernetes "converts" its services and pods/endpoints into a ServiceEntry through Istio .   A service with two pods and two endpoints -  10.244.3.1, 10.244.3.2 -,  will be mapped to a ServiceEntry - host / IP address endpoint -. Endpoints  represent the IP addresses that a service sends its requests to. The "hosts" in the  ServiceEntry   Yaml file are associated with the service. A ServiceEntry adds newly added services to the service registry - database of existing services in Istio -, this allows existing services to be able to access these newly deployed ones. Below is an example of a service its endpoint and the resulting service entry : We have two nginx pods and the service entry that allows access through the host " service1.default.svc.cluster.local " to  the nginx applications.
Read more

Kubernetes 101 : - Scaling a Stateful set -

Image
The below diagram represents the scaling down of a  statefulset named  db1 : Before scaling down our statefulset , we check the pods and we see that there are three replicas: We can tell from the names of our pods that they are managed by a  statefulset is named  db1 . We look for that statefulset in our system using the below: Then we can scale out statefulset down to two pods using the below command: We check to see if the statefulset was successfully scaled down to two pods using the following: Remark: The " --record " parameter created an annotation as we see using the " kubectl describe " command:
Read more

Python 101 : Generators and Lists

Image
A list is a collection of val ues stored in memory. A generator does not build a list, it only tells us how to generate these values with nothing stored in memory. Our examples below use the " range() " iterator and they produce the same results. Below is our list example: And our generator example: The generator expression generate values only when they are needed which saves us system resources, especially with huge amounts of data.
Read more