Review of our Google Cloud Zürich Meetup #4

From Containers to Serverless with Cloud Run

Jakob Ehrl, Customer Engineer of Google Cloud Zürich, introduced us into his sandbox experiment with Cloud Run. Due to the current home office situation, he can’t travel to Google’s offices in Zürich by bike (15km). Together with his girlfriend, he decided to run every second day and log the activities to Strava. In order to play around with the gathered data, he shows in his uncommon painted presentation and lots of interaction with the audience, how the GPS data points from his watch get into BigQuery where he can analyze them. He presents a pipeline of entirely serverless building blocks including

Google Cloud Storage (ingress and intermediate processing results)

PubSub (for service communication)

Cloud Run (for parsing the input data in a given Java application)

Cloud Functions (as an alternative to Cloud Run)

BigQuery (as data warehouse)

Data Studio (for visualization)

He said it took him about a week to put the things together by mostly clicking on the Google Cloud Console – without counting the nights!

Typical Kubernetes User Questions

(The numbers in brackets refer to mm:ss in the recording below)

The 45min Q&A session about typical Kubernetes user questions with nine Kubernetes Engineers Sebastian Nickel and Cyrill Troxler started with survey questions. It showed that most of the attendees already have Kubernetes experience or are evaluating Kubernetes. They are using Raspberry Pi Clusters, Kube Flow or Minikube to mention a few.

(3:05) What are the best practices making services “elastic” to scale up and down automatically?

(5:05) What are the experiences with preemptible nodes with GKE?

(8:20) What is the best way to create resources in Kubernetes?

(11:48) What about vertical scaling nodes?

(13:23) Do you work with customer resource definition?

(13:38) Why do you not use StackDriver for logging?

(17:22) How to backup the Kubernetes cluster? What are your experiences with persistent volumes and backup?

(19:13) Do you have hints or experiences in config pod eviction policies?

(20:20) How to create/use infrastructure around Kubernetes in the cloud (databases, caches, etc.)?

(21:58) You don’t use databases within Kubernetes?

(22:50) What/how should we monitor in Kubernetes?

(24:50) Do you monitor outside Google infrastructure?

(26:15) Do you provide a cost driven approach to your customers or are they just paying for Google Cloud resources?

(27:10) How do you make sure that the load is equally distributed between zones?

(29:40) Do you have customers with hybrid setups? What about VPNs with i.e. 3 GBits of traffic?

(33:52) How is kubectl pronounced?

(34:52) Is Kubernetes secure?

(38:23) Is it hard to run an own Kubernetes cluster like an on premises one?

Further outlook with the new online experience

Our next Google Cloud Meetup is planned for August 25, 2020 – hopefully in Zürich City again. With the very good experience doing an event online, but the lack of networking capabilities at the end, we want to combine the advantages of both, online and offline, as soon as it is allowed again. If you have an interesting project running on Google Cloud and think it could be interesting for others, feel free to contact us! We also very much appreciate suggestions for topics and possible speakers, by either meet-up-message, by using the contact form on our website, or via email to one of your contacts here at nine. We would like to keep things exciting. We look forward to seeing you again!