Note: This blog has two authors.

What is context caching?

Vertex AI is a Google Cloud machine learning (ML) platform that, among other things, provides access to a collection of generative AI models. This includes the models known under the common name “Gemini models”. When you interact with these models you provide it with all the information about your inquiry. The Gemini models accept information in multiple formats including text, video and audio. The provided information is often referred to as “context”. The Gemini models are known to accept very long contexts.

Production systems are being monitored for reliability and performance tracking to say the least. Monitored metrics ‒ a set of measurements that are related to a specific attribute of a system being monitored, are first captured in the executing code of the system and then are ingested to the monitoring backend. The selection of the backend often dictates the methods(s) of ingestion. If you run your workloads on Google Cloud and use self-managed Prometheus server and metric collection, this post will help you to reduce maintenance overhead and some billing costs by utilizing Google Cloud Managed Service for Prometheus for collecting and storing Prometheus metrics.

If you do not care about how much money you spend on your Google Cloud project then this post won’t be interesting to you. If you do, you might find the information below useful.

Google Cloud documentation gives examples of the automated cost control responses. It describes an option to send a notification message to PubSub in addition to the email each time the billing budget alert is triggered.

Last updated: August 12, 2024.

Everyone today heard or read about GenAI, ChatGPT and other AI things. There are a lot of terminology, abbreviations and other clever words. I found myself troubled to remember all of them. So I decided to write down my definitions for each of these terms.

TL;DR; If you are familiar with AI terminology, you might want to stop reading this post here. For the rest of the readers, welcome to my personalized thesaurus of GenAI terminology.

This article surveys various health checks in Google Cloud. If you want to learn more, leave your preferences in the feedback form.

Generally speaking, a health check is a function or a method to indicate a general state (a.k.a. health) of the underlying service. Some products elaborate the definition of “general state” to be something particular, such as the ability of the service to respond to requests.

Health checks are an important instrument of service observability. When provided as a tool or service they replace development of coding metrics ingestion, collection and analysis. Health checks often come integrated with alerting or incident response solutions. In many scenarios health checks can be sufficient to increase product reliability to desired level.

Google Cloud lets you run Kubernetes in three flavors:

• Vanilla is when you do all on your own. This is also the quickest “lift and shift” strategy to migrate your cluster to cloud. Essentially it is just a group of virtual machines that run on Google Compute Engine (GCE).
• Managed that shifts administration and maintenance tasks from DevOps teams to the cloud service. See Google Kubernetes Engine (GKE) Standard cluster architecture and Autopilot for more details.
• Knative, that is sometimes referred to as cloud native, which hides control plane and other infrastructure details behind the familiar interface of workload launching. Cloud Run offers running service and job workloads using the GKE platform behind the Knative interface.

Many DevOps teams prefer the managed flavor to enjoy a balance between carefree administration and the level of control that is very close to vanilla Kubernetes. Comparing GKE Autopilot and Standard, many prefer Standard due to higher control granularity over node management, security and version configuration and other options. In the cluster observability domain, these differences are less distinctive since both come with a rich set of monitoring and logging capabilities including control plane metrics.

You may have seen this notice when opening SLOs Overview in Cloud Console.

This notice announces a recent change in the way of defining services for Cloud Monitoring. Before the change, Cloud Monitoring automatically discovered services that were provisioned in AppEngine, Cloud Run or GKE. These services were automatically populated in the Services Overview dashboard. After the change, all services in the Services Overview dashboard have to be created explicitly. To simplify this task, when defining a new service in UI you are presented with a list of candidates that is built based on the auto-discovered services. The full list of the auto-discovered services includes managed services from AppEngine, Cloud Run and Istio as well as GKE workloads and services. Besides UI you can add managed services to Cloud Monitoring using the services.create API or using the Terraform google_monitoring_service resource.

Google Cloud supports service monitoring by defining and tracking SLO of the services based on their metrics that are ingested to Google Cloud. This support greatly simplifies implementing SRE practices for services that are deployed to Google Cloud or that store telemetry data there. To make it even more simple to developers, the service monitoring is able to automatically detect many types of managed services and supports predefined availability and latency SLI definitions for them.
When you define a new SLO you are prompted to select a predefined SLI or to define your own.

I wanted to introduce a support for “series” ‒ an ordered collection of posts about the same topic. For example, this post is the 4th post in the series “Getting Started with Hugo”. And I wanted to enable navigation between different posts from the same series instead of just going to “next” or “previous” post in the blog. I looked on internet and found many articles about how to do it using Hugo. The posts that I found used older versions of Hugo because of configuration parameters and methodology that they used. I learned from these posts and came with my own implementation.

Hosting

After a brief review of the selected Hugo settings, let’s review the decisions about hosting. I used only two criteria when I was choosing hosting for my blog website:

• Convenience – the service should be simple to use, allow automation, to have both CLI and Web interface and, ideally, familiar
• Costs – it should be cheap or, ideally, it should be free

The choice for the website source code management was straightforward. Github answers both criteria. It is simple. Supports automations via workflows and actions. And I work with it everyday. Needless to say that it is free (for now).