Interview Experience in Fresh startup EDJX
concepts how web request is going. DNS resolution, if not stored in local cache then global resolution to check for dns, once mapping is done then it will hit server.
Do you know about A, AAA, CNAME records? Have you used Route53?-> (I used but I was trapped here, my problem has been always not to see things too deep), A record for ipv4 and AAA for ipv6 (I remember I have created this DNS mapping during very first )
Will it go to load balancer?-> Yes
What do you think in case will there be a single load balancer or multiple, lets say I am hitting Facebook.com ? -> there will be multiple. I got the discussion to Cloudfront , region wise we will split our traffic to various locations (I know company is based on edge computing), from there we can set up load balancer and autoscaling groups.
Lets say you request has reached load balancer? After that? We can use service like nginx or haproxy? -> Sir request will go to apiGateway, do we need to use apiGateway here? (I paused), no sir here the ingress service is itself acting like apigateway here? It will then go to various services like cluster, nodeport. I said like I have used the service haproxy and keepalived in Nokia, where keepalived was used to map virtual ip to physical ops and also setup priorities there.
ClusterIP or nodeport? -> Nodeport sir! (I flunked), while using nodeport is outside clusters, clusters can be used within the cluster as well. From ingress, the service can go to cluster ip. What sort of, type of components will be there within clusters?-> He meant by deployments and stateful sets. -> Stateful sets are created keeping session in state, deployments are an abstraction for one or multiple pods, these multiple pods can be replicated in case more traffic is coming
But then we can use HPA as well, then why we need deployments?
Yes me answer this, deployment is full state which have pods, configmaps, no of replicas which is responsible for overall deployment of single unit of functional element. DO you agree?-> I said yes as I remember deployments have replicasets and these replica sets ensure minimum no of nodes are existing. After this request may go to
Have you used Db inside or outside the cluster?-> I said we used both. Redis as an inside cluster and also outside, gave examples of volume sharing of docker containers
What all monitoring you have used? Grafana or prometheus, what all data points are there? -> We used Grafana to see our inventory, and here we used sumologic and I have used AWS cloudwatch
What is difference between list and tuple? -> Objects in tuple are immutable, they can’t be altered, once created a new object will not override the existing object.
How will you develop like say you have one app in developer. So
GitHub actions-> no sir, barely started.
How a linux machine gets boot up?->
Well we use github actions extensively here. -> I have used Jenkins to integrate code
What is the purpose of helm?-> It is like a package manager for kubernetes. So if we want to create service accounts, clusters, clusterrolebindings, deployments , services all together , so one thing is we create. All the stuff manually, but this would be a tedious task and so we can bundle all the things together to create a single resource and with firing 1 single cmd we can run all the resources together.
Terraform-> again knew about cmds > terraform init, plan, apply . Have you created any chart? Honestly didn’t got any opportunity since I was good in python hence used boto3 module, also used Cloudformation and Heat stack for template.
Which all modules have you sed in python apart from boto3 have you done OOPs programming in python as well or ? -> boto3, openpyxl (for excel operations in python)
Any experience in Golang since we use Solang over here-> not much but have worked in java