Thank you for > 10k Subscribers ??

You are over 10k Subscribers on this newsletter at the moment of sending this issue (10245 to be precise). When I started this from a post on X last year I never imaged we will be here ??

Thank you reading, following, subscribing and DM'ing me suggestions. I will do my best to provide more value. Over the next few weeks I will send out a survey I hope you will take a time to participate in ????.

In case you did not take your ticket for KubeCon North American in Salt Lake City in two weeks from now. I have some discount codes you can use:

All-Access (ABDEL_AAI $899 Individual, ABDEL_AAC $1,549 Corporate)

KubeCon Only (ABDEL_I $600 Individual, ABDEL_C $1,250 Corporate)

The News

GKE

• GKE Gateway Controller New release: Release 2024-R2 of the Gateway Controller (the components in GKE that allows you to use the Gateway API to expose your apps to the internet) brings a lot of new features.
• Use NVIDIA H100 80GB GPUs with new machine types: Support has been added for a3-highcpu-1/2/4g. These machines also support Dynamic workload scheduler, Spot VMs and Spot Pods in GKE Autopilot.

Google Cloud

• New version of Google Cloud SDK (gcloud): Version 498.0.0 with many improvements is available.
• Cloud Run deploy to multiple regions: Cloud Run added support (In Preview) to deploy a service to multiple regions with a single API Call (gcloud run deploy) instead of calling the API to deploy to each region individually. This is useful if you want to serve traffic from multiple regions for HA and latency purposes. Keep in mind you still need to configure a loadbalancer yourself.
• Artifact Registry support for custom policies: You can define organization and custom policies on Artifact Registry. This allows you to enforce policies like Customer Managed Encryption Keys (CMEK), metadata…etc

The Editorial

GKE

• Reference Architecture - Github Runners on GKE : This new reference architecture guides you through the process of setting up self-hosted GitHub Actions Runners on Google Kubernetes Engine (GKE) using the Terraform.
• Benchmark app performance with Locust on GKE : Learn how to benchmark an application performance on GKE using Locust load testing. This way you can ensure you application will perform under load.
• Tweak HPA on GKE for Inference Workloads : Autoscaling Inference workloads is tricky. Since these are not traditional workloads and they emit different metrics, knowing how to properly setup the Horizental Pod Autoscaler to properly scale up and down based on utilization is important. Luckily this guide has some tips and tricks.
• GKE Workload Identity is now named Workload Identity Federation : What changed? This article goes into details.
• Running Ray on Google Cloud: This is the first of a series of 4 articles covering Ray, how to deploy it on GKE and how to use it to train a model.

Google Cloud

• Using Cloud SDK Library JSS : Jump Start Solutions are a series of templates published by my team which includes end to end solutions that tackles a specific use case using Google Cloud Products. The team added a new solution to demonstrate how Client Libraries can be used to Interface with various parts of Google Cloud.
• GitLab on Google Cloud : In a landing page with links to documentation, tutorials, guides and best practices for running Gitlab on Google Cloud.
• Cassandra to Spanner Proxy : Is a proxy that allow you to connect apps that usually talk to Apache Cassandra databases to Spanner.
• Cloud Deploy Rollout Retries : Cloud Deploy (Google Cloud CD Service) supports a rollout retry. if a rollout fail you can set it to retry a certain number of times before it rolls back.
• Deny Policies and why they are important : We are all familiar with allow policies (both in IAM and firewalls) but what about Deny policies. A way to tell Google Cloud to not allow certain actions to happen regardless of the permissions a principal has. This article covers some use cases where deny policies can be useful.
• Database Center: I have read about this a while ago. In Preview now it’s an GUI which is a central pan of glass for everything databases in Google Cloud. And has Gemini built-in.