Autospotting on AutoSpotting

Why I Just Declined a Job Offer from a Billionaire Friend

Sat, 01 Jun 2024 00:00:00 +0000

Why I just declined a job offer from a billionaire friend

June 01, 2024

This is an expanded version of a LinkedIn post I wrote the other day.

The first email I read this week was this one I received from a billionaire friend, who wanted to talk to me on very short notice:

He’s the founder and CEO of a pretty large and well known company he still largely owns, and I respect him a lot so I won’t reveal his identity.

New AutoSpotting Bugfix Release - 1.3.1

Mon, 29 Apr 2024 00:00:00 +0000

Fixing regressions in 1.3.0: incompatibility with instances out of the DefaultVPC and Lambda timeout when replacing Spot instances

April 29, 2024

Hello,

After releasing AutoSpotting 1.3.0 a couple of weeks ago, I’ve been working with a few customers who noticed a couple of regressions introduced in this latest version:

AutoSpotting failed to launched instances except in the Default VPC (major).
AutoSpotting Lambda timeouts while waiting for terminated Spot instances to be in Running State in case the Grace Period is longer than 2 minutes (medium), resulting in failure to clear suspended ASG. processes**.**

This release is meant to address those and is strongly recommended to users that don’t use Default VPCs for EC2 instances launched by AutoSpotting.

New AutoSpotting Releases

Thu, 11 Apr 2024 00:00:00 +0000

After almost 8 months since the previous AutoSpotting release, we’re happy to announce not one, but two AutoSpotting releases at the same time, packed with lots of improvements.

I was hoping to release this much earlier, but somehow each time we got quite close, customers reported yet another little but important edge case that would be be great to fix and then another and so on…

This finally resulted in a few dozens of relatively small reliability, correctness and performance improvements addressing lots of edge cases throughout the code base.

New AutoSpotting Release: Mixed Autoscaling Groups Support

Tue, 04 Jul 2023 00:00:00 +0000

How ECS capacity providers inspired me to fix an ancient AutoSpotting issue

July 04, 2023

I’m excited to announce that a new version of AutoSpotting is now available.

For some context if you’re not familiar with AutoSpotting, it’s a tool that makes it easy to adopt Spot instances in existing Autoscaling groups, without requiring configuration changes, by replacing their instances with Spot clones using attach/detach API calls.

For more information about AutoSpotting you can have a look at AutoSpotting.io or check out our Open Source code on GitHub.

AutoSpotting Comparison Matrix

Mon, 15 May 2023 00:00:00 +0000

Over the years many people asked us to explain the difference between AutoSpotting and standard Autoscaling groups, as well as between AutoSpotting Community edition and the commercial offering on the AWS Marketplace.

We just added this feature comparison matrix on GitHub, so you can easily see how they all compare.

Let us know if you have any further questions.

-Cristian

Keep reading

[

Another New AutoSpotting Release in Less Than a Week

Sat, 15 Apr 2023 00:00:00 +0000

Fixing ECS load balancer draining

April 15, 2023

Over the last 7 years I’ve been working hard to make AutoSpotting the best solution for running Spot Instances in Autoscaling groups.

Just a few days ago I released our biggest release ever which makes AutoSpotting managed groups of Spot instances even more resilient to low capacity situations than the vast majority of Autoscaling groups configured with a single OnDemand instance type.

As I always keep improving the experience based on user feedback, I just released a new version of AutoSpotting that improves reliability for ECS users, making AutoSpotting with ECS a more highly available solution than plain ECS itself!

LeanerCloud Product Updates

Wed, 12 Apr 2023 00:00:00 +0000

After less than 7 weeks from our previous AutoSpotting release, we’re happy to announce a new one, our biggest release so far in the entire 7 years history of AutoSpotting.

It brings over a dozen improvements in performance, reliability, correctness and ease of use, including the following major enhancements:

Proactive Spot instance launch on Spot terminations*, reducing the instance replacement time by avoiding the use of temporary On-Demand instances.*
Diversified failover to On-Demand instances during periods of high-demand of EC2 capacity provides even more resilience to Insufficient Capacity Errors(ICE) than the typical AutoScaling groups configured with a single instance type.
Automatically determine the Spot product for each AutoScaling group*, resulting in a simpler setup, correct instance type selection, accurate billing and higher diversification for Windows, RHEL and Suse Linux Spot instances.*
Supporting installation of AutoSpotting in any AWS region as its main region, reducing the complexity and networking costs of the previous cross-region setup.
Fixed a number of issues related to Security group configuration*, which sometimes caused Spot instances to fail to launch or start with incorrect Security Groups.*

This updated version of AutoSpotting is available on the AWS Marketplace and can be installed as usual using CloudFormation or Terraform.

Weekly Update - 17 Mar 2023

Fri, 17 Mar 2023 00:00:00 +0000

Hello and welcome to this week’s status update.

Let’s dive tight in!

AutoSpotting Architecture changes as result of my current customer onboarding

This week was another one largely focused on AutoSpotting development, most of the work resulting from the conversation I had with the relatively large enterprise customer I started to onboard last week to AutoSpotting.

This is high priority work for me, since their PoC alone would be bigger than all my other customers combined, so I did my best to make it work for them.

Weekly Update - 13 Mar 2023

Mon, 13 Mar 2023 00:00:00 +0000

Hello and welcome to my latest weekly progress report.

I’m a bit late to this week’s update, I normally do these on Fridays as a way for me to reflect on the previous week, but on Friday I was busy helping a new potentially large AutoSpotting customer get started.

That didn’t leave me enough time to write this, but I’ll talk about the opportunity and the work itself towards the end of this progress report.

Weekly Update - 3 Mar 2023

Fri, 03 Mar 2023 00:00:00 +0000

Hello and welcome to this week’s progress report.

This is also a bit of a special one, since it marks more than 6 months since I left AWS to double down on AutoSpotting development.

Without further ado, let’s dive right in!

Terraform updates

I started the week by working on bringing the AutoSpotting Terraform module in sync with the CloudFormation template, after previously it was rolling out an ancient release of AutoSpotting from October 2021, with lots of known issues.

Weekly Update - 24 Feb 2023

Fri, 24 Feb 2023 00:00:00 +0000

Hello, and welcome to this week’s update.

Without further ado, let’s dive right in.

Little recap from last week

In my last week’s update I mentioned how I’ve been working most of the week helping a new customer onboard.

They were reluctant to run the latest AutoSpotting version because of their blue-green deployment strategy, but they wanted to try an older version that still works in the legacy cron mode. They were initially very happy about the way I supported them to use an older version, and even said it was the best support they ever got from a vendor.

Weekly Update - 17 Feb 2023

Fri, 17 Feb 2023 00:00:00 +0000

Welcome to this week’s progress update.

This was a shorter week than usual, as Monday had some family emergency and largely took the day off, but it was otherwise a pretty productive week, with many things moving forward and a lot of things to write about, so this is probably my longest update so far, but more technical than usual.

It was again a week focused on development work, worked on multiple things but still on the same themes of reducing friction for new users and making life easier for existing users.

Weekly Update - 10 Feb 2023

Fri, 10 Feb 2023 00:00:00 +0000

Hello, and welcome to another weekly update.

(I forgot to publish this last week here on beehiiv, so I’m now publishing it together with the current week’s update. The current week’s update will be scheduled 12h later, in order to accommodate beehiiv restrictions, sorry about that.)

LeanerCloud GUI

This week I continued the development of the new LeanerCloud GUI.

I made a lot of progress, the GUI is now able to estimate the savings, updating the estimates based on the selected ASGs and their configuration, and also can write a few of the commonly used tags, as well as reading the values set in the previous runs.

Monthly Update August 2022 - Liftoff

Sat, 10 Sep 2022 20:36:12 +0100

Hello,

I’m happy to announce that I finally left my full-time job last week, so this is my first status update as self-employed. Over the last few weeks I’ve been vacationing a lot So far there’s not much to share

It’s been a while since my previous update in here, which was at the end of October.

I’m now trying to resume these monthly updates, so in this first post after this gap I’d like to give a few updates covering the last months, and also my plans going forward.

Monthly Update October 2021

Sun, 31 Oct 2021 20:36:12 +0100

Hello,

As usual, here’s my latest monthly report.

Going forward I’ll release these reports here on the CloudUtil blog and I’ll only share the link through my mailing list and our new Slack #announcements channel.

If you want to be notified when I write a new one, don’t forget to sign up to the CloudUtil newsletter shown at the bottom of the page, or join us on Slack.

I also published as blog posts most emails I had sent to the AutoSpotting newsletter over the last year, so you can see how it changed over this time. I had to edit them a bit to fix URLs and clean them up a bit, bear with me in case there were any issues.

Monthly Update September 2021

Mon, 04 Oct 2021 03:36:28 +0100

Hello,

As usual, here’s my latest monthly report.

AutoSpotting

Since the previous month there hasn’t been much going on in the mainline repo, at least not from my side. The only thing I did there was updating dependencies, including instance type information for the recently released M6i instance types.

From the community side, I’m very thankful to mello7tre who contributed a couple of significant changes:

allow us to reuse existing Spot instances instead of always launching new ones, saving some time when doing Spot replacements in certain situations.
added a global configuration flag (defaults to off) that allows us to disable the Instance Rebalancing Recommendation events, which sometimes get emitted without resulting in a Spot termination, creating additional instance churn. The caveat is that with this enabled you only get the 2-minute termination notification, and don’t get to run the instance termination lifecycle hooks, but it gives you a bit less instance churn if you don’t need those.

This month I also updated the Terraform code to match CloudFormation and documented it as officially supported installation method on the AWS Marketplace.

Monthly Update August 2021

Sun, 29 Aug 2021 10:13:42 +0100

Hello,

As usual, here’s my monthly progress report, this time covering August 2021.

AutoSpotting

As I mentioned in my previous report, I’ve been working on releasing the new AutoSpotting stable build on the AWS Marketplace. After many months of hard work, I’m happy to announce that the release finally landed on the Marketplace a few days ago, and it’s probably the first serverless paid product built with Lambda available on the Marketplace.

Monthly Update July 2021

Mon, 09 Aug 2021 07:24:54 +0100

Hello,

Another month, another community update. This one comes about a week late as I’ve been on a long vacation and some things had to give.

Progress report - July 2021

AutoSpotting - throughout my vacation I’ve been working a lot on a still unreleased branch that prepares AutoSpotting for being published on the AWS marketplace as a paid Docker product, charging a small percentage of the savings. This was way harder than I expected, as AutoSpotting is probably the only Lambda-based software being made available as paid Docker product so a lots of things had to be figured out and tweaked for making it work under the Marketplace constraints. There are many not-so-functional changes under the hood, such as calculating the generated savings, repackaging the Lambda to use Docker images in order to fit the Docker product requirements of the marketplace, charging a percentage of the savings amount through the Marketplace APIs from a new Fargate task(don’t ask! ;-) ), etc. I’ve learned a lot while building this(I’ll probably publish a technical blog post about it at some point) and it was a lot of fun and even though it basically works, I still have a number of things to figure out or update(especially docs). My goal is to have this published by the end of August, fingers crossed! I’ll do my best to not cause much impact to the OSS users, but some things might be rough on the OSS side for a while after this code lands.

Monthly Update June 2021

Sun, 04 Jul 2021 11:54:03 +0100

Hello,

Another month, another AutoSpotting community update.

Progress report - June 2021

EBS-Optimizer - To be honest this month I haven’t made as much progress on AutoSpotting as I had hoped, but I started a couple of new workstreams which kept me busy. The first is EBS-Optimizer, my new tool dedicated for EBS volume upgrades from GP2 to GP3, resulting in up to 20% better cost and often better performance than GP2 volumes.

Monthly Update May 2021

Wed, 02 Jun 2021 11:40:52 +0100

Hello,

As promised in my last monthly email, I’m now sharing monthly the progress achieved in the previous month and what’s coming soon.

I’ve also resumed the monthly AskMeAnything-style community calls, and below you can see the recording of our previous one.

(The recording of this call is unfortunately no longer available online.)

Progress report May 2021

I helped polish further the contribution by Alberto (mello7tre on Github), which simplified the event handling logic by introducing a FIFO SQS queue for serializing AutoScaling operations (such as temporarily extending the maximum capacity and suspending AutoScaling processes) instead of previously calling an auxiliary Lambda function with concurrency 1. The code and internal architecture is significantly cleaner and much more robust after our initial tests and I’m very happy with the way it works. I’d really appreciate if you could give it a try in a test environment and please let me know if you run into any issues.
this month I also merged the EBS storage upgrade logic that converts GP2 volumes to 20% cheaper and more performant GP3 volumes for EBS volumes smaller than 170GB. There’s also a change for similarly converting IO2 volumes to IO3, but that’s unlikely to make much of a difference for Spot use cases. This is also available in the current nightly build, and here’s how it looks like for GP2 volumes:

Monthly Update April 2021

Sat, 01 May 2021 21:48:07 +0100

Hello,

It’s been a while since the previous status update and also I really missed seeing some of you like I did in our monthly webinars.

I decided to give a status update for the last few months and considering it’s the beginning of the month going forward I’ll do my best to do this on a monthly basis, sharing the progress since the previous month and what’s coming soon.

Monthly Update October 2020

Fri, 16 Oct 2020 21:54:24 +0100

Hello,

I’m happy to announce that the massive code change that has been in development since more than a year ago has been merged and after a few further improvements it’s now ready for Alpha/Beta testing by a wider audience.

What’s new?

Event-based instance replacement

In a nutshell, when deployed using the latest CloudFormation infrastructure code, AutoSpotting is now reacting to all new instance launches, in addition to the legacy cron-based execution mode that’s still available as before. Under normal circumstances newly launched on-demand instances are now immediately replaced with spot clones, avoiding to run the spot clone in parallel to the AutoScaling group they were launched for which has always been a quirk of AutoSpotting when compared to its alternatives.