Aurora Serverless v2's Broken Promises

Corey begins by reciting the classic poem “The Aurora Serverless Road Not Taken” and telling the woeful tale of AWS’s diverging Aurora Serverless offerings. He gives a brief history of the two versions and highlights the issues with version 2, namely that it doesn’t actually meet the definition of “serverless,” doesn’t scale to zero, and comes with frustrating extra charges.
00:15 Advertisement - Cribl - snark.cloud/cribl-splunk
01:33 Beatnik Corey recites the classic poem “The Aurora Serverless Road Not Taken”
02:35 Aurora Serverless v1 (2017), and Aurora Serverless v2 (2022)
04:04 Corey talks about the problems with Aurora Serverless v2, namely that it doesn’t scale to zero
05:49 v2s use of Amazon RDS proxy and ANOTHER extra charge
- Never miss an episode -
► Join the Last Week in AWS newsletter: www.lastweekinaws.com/
► Subscribe wherever you get your podcasts
- Help the show -
► Share your feedback: www.lastweekinaws.com/feedback/
► Subscribe wherever you get your podcasts
- What's Corey up to? -
► Follow Corey on Twitter (@quinnypig): / quinnypig
► See our recent work at the Duckbill Group: www.duckbillgroup.com/clients/
► Apply to work with Corey and the Duckbill Group to help lower your AWS bill: www.duckbillgroup.com/services/
This episode is sponsored by Cribl Learn more here: snark.cloud/cribl-splunk

[@rodrigomataibanez8859]

Apr 2023 and this video still tells the truths about serverless v2

[@kevindunlap2832]

Redshift serverless…. The saga continues!

[@hoohaw286]

Jesus. So much pimping of products before it begins and then nothing but clickbait lies and stupidity. Let's take a medium size prod database non serverless db.r5.2xlarge instance type with 64GB of memory. $2/hour. To run that nights, weekends when things are slow with serverless v2 takes you down to 0.5 ACU at 6 cents an hour??? It is far from "buried" in the documentation it is up front on every page including pricing page and every single AWS youtube video mentions it. Also are you really that discombobulated that there is a v1 and v2? Your tech life must be a hellish nightmare of confusion and anxiety if you've never heard of versioning before. The implicit promise of serverless is not having to manage real or virtual infrastructure or configure and manage auto scaling settings. Not about scaling below 1GB at 6 cents an hour. Also AWS has a long history of transparently rolling out improvements so they will most likely address even that at some point in future. This video is pure half assed click bait.

[@jaysistar2711]

Our problem is traffic spikes during weekdays. It will scale to 22 ACU, then down to 0.5 ACU. This actually saves us quite a bit of money on the weekends by not having a big machine running all the time.

[@SeanNelsonCesa6]

This is my first time seeing content from this channel, and I don't think I'll seek out more. This is nitpicky complaining that seems to be written from the perspective of someone running a tiny application, like a blog, serving dozens of people per day, with a specific compute structure - largely Lambda instances hitting RDS directly. Many AWS-focused professionals are worried about maintaining applications used by thousands minimum, with a more controlled compute and connection strategy.
It seems to me that Serverless v2 is designed for those applications that are large enough that it can never be scaled to zero anyway. So, maybe a workload of .5 on a night + holiday, and 20 on a peak workday. Often, these requests are not coming from a flood of Lambda requests, but from a small pool of application servers, which may themselves be serving APIs ingested by Lambda or other compute resources. If this is your model, Aurora Serverless v2 seems like a much, much, much more cost effective option than maintaining Multi-AZ m5.2xlarge running 24/7.