etcd VS consul

GitHub Repository. etcd-io/etcd

etcd (https://etcd.io/) is an open-source leader-based distributed key-value datastore designed by a vibrant team of engineers at CoreOS in 2013 and donated to Cloud Native Computing Foundation (CNCF) in 2018. Since then, etcd has grown to be adopted as a datastore in major projects like Kubernetes, CoreDNS, OpenStack, and other relevant tools. etcd is built to be simple, secure, reliable, and fast (benchmarked 10,000 writes/sec), it is written in Go and uses the Raft consensus algorithm to manage a highly-available replicated log. etcd is strongly consistent because it has strict serializability (https://jepsen.io/consistency/models/strict-serializable), which means a consistent global ordering of events, to be practical, no client subscribed to an etcd database will ever see a stale database (this isn't the case for NoSQl databases the eventual consistency of NoSQL databases ). Also unlike traditional SQL databases, etcd is distributed in nature, allowing high availability without sacrificing consistency.

etcd is an open-source leader-based distributed key-value datastore designed by a vibrant team of engineers at CoreOS in 2013 and donated to Cloud Native Computing Foundation (CNCF) in 2018. Since then, etcd has grown to be adopted as a datastore in major projects like Kubernetes, CoreDNS, OpenStack, and other relevant tools. etcd is built to be simple, secure, reliable, and fast (benchmarked 10,000 writes/sec), it is written in Go and uses the Raft consensus algorithm to manage a highly-available replicated log. etcd is strongly consistent because it has strict serializability, which means a consistent global ordering of events, to be practical, no client subscribed to an etcd database will ever see a stale database (this isn't the case for NoSQl databases the eventual consistency of NoSQL databases ). Also unlike traditional SQL databases, etcd is distributed in nature, allowing high availability without sacrificing consistency.

Unfortunately there is a big problem with custom controllers, they can't handle huge amount of data for several reasons. Kubernetes relies on ETCD for all data storage, which limits scalability, flexibility, and performance for complex or high-volume workloads. What kind of problems I'm talking about?

In Amazon EKS, AWS Config helps you track key components such as: EKS control plane logging, VPC settings and network exposure, encryption status for logs and secrets and IAM roles used by worker node groups. It can detect misconfigurations like: 🚫 Publicly accessible EKS clusters ⚠️ Disabled encryption for secrets stored in Kubernetes ETCD ⚠️ Ensures EKS clusters are running on currently supported versions,

curl -LO https://github.com/etcd-io/etcd/releases/download/v3.5.21/etcd-v3.5.21-linux-amd64.tar.gz tar xzf etcd-v3.5.21-linux-amd64.tar.gz

Etcd is a distributed key-value store, somewhat like Redis, but it operates quite differently under the hood (more on this later). It's implemented in Golang and is fully open-source.

> https://github.com/etcd-io/etcd/issues/9771

> stale bot marked this as completed (by fucking closing it)

Ah, yes, what would a Kubernetes-adjacent project be without a fucking stale bot to close issues willy nilly

Using a service discovery mechanism: A service discovery mechanism, such as etcd or ZooKeeper, can help to manage the complexity of microservices by providing a centralized registry of available services and their instances.

ETCD_VERSION='3.5.4' wget https://github.com/etcd-io/etcd/releases/download/v${ETCD_VERSION}/etcd-v${ETCD_VERSION}-linux-amd64.tar.gz tar -xvf etcd-v${ETCD_VERSION}-linux-amd64.tar.gz && cd etcd-v${ETCD_VERSION}-linux-amd64 cp -a etcd etcdctl /usr/bin/ nohup etcd >/tmp/etcd.log 2>&1 & etcd

Consul, HashiCorp's distributed system tool, isn't just a one-trick pony for service discovery—it's a Swiss Army knife that also handles configuration management and network segmentation.

I had a distaste for Nomad before the rug pull[1], but hopefully no serious person would voluntarily choose Nomad after that stunt. Doubly so that going with the Nomad ecosystem mandates self-hosting as one cannot $(aws elastic-nomad create-cluster). I was similarly surprised that HashiCorp didn't even offer "you pay for compute but we will administer your control plane" since both the "community" and the "enterprise" links from the front page both say "self-managed"

I will be super curious to see if IBM Cloud actually does ship $(ibm-cloud-cli create-nomad)

1: I also didn't realize they rug pulled Consul, too; that's just cruel https://github.com/hashicorp/consul/blob/v1.20.5/LICENSE

No need for NGINX reloads: Since NGINX queries the Consul Go API client for healthy services on each request, there’s no need to reload NGINX whenever a service moves between nodes or when new instances are added.

When a backend starts or stops, something needs to update, whether it’s Consul, kube-proxy, ELB, or otherwise. To stop a worker without incurring failures, you need to prevent the load balancer from sending new requests and then finishing existing ones.

Consul - To set up secure network communication via mTLS, service location and certificate issuance

Hashicorp Consul

The JHipster scaffolded sample application has a gateway application and two microservices. It uses Consul for service discovery and centralized configuration.

Service Discovery: Microservices need to discover and communicate with each other dynamically. Service discovery tools like etcd, Consul, or Kubernetes built-in service discovery mechanisms help locate and connect to microservices running on different nodes within the infrastructure.

After some research and testing, I landed on using Consul and Fabio as the demo infrastructure. Of course, there is a myriad of other options to accomplish this task, but because of the low configuration and ease of use, I was impressed with this pairing. Both projects are mature and well-supported, and very flexible--just because you can run them with low configuration, doesn't mean you have to. I wanted to keep this demo constrained, but the exercise did get me excited about exploring things further: circuit breakers, traffic splitting, and more complex service meshes.