https://serverless-stack.com/ logo
Join Slack
Powered by
<https://twitter.com/MarcJBrooker/status/151719381...
# random
t
r
Such a big deal if it delivers on it's promises.
Interesting thing mentioned in the docs but I didn't see mentioned in the announcement: a flavor of RDS Proxy specifically optimized for Aurora Severless v2
f
Does v2 not support Data API?
And doesn’t scale down to 0 from I’m reading on Twitter
a
Yeah, no data api, no scale to 0, also weirdly no VPC requirement. And twice as expensive as v1, so the minimum monthly cost per cluster is ~$45-50.
t
“serverless”
t
Wait no data api
Wtf
r
Oh, I didn't realize Data API wasn't supported!
Weirdly enough, RDS Proxy is supported
No scale to zero is def the big stinker for me, but I'm weirdly pretty happy with it otherwise
l
Wait, what?
I'm reading the official memo and one of the use cases is actually a fleet of auroras that scale down to zero when not used
t
I guess RDS Proxy is similiar to data api, can put it outside vpc and secure with iam
yeah that's odd
l
Sooo expensive
Has anyone here tried out Fauna or HarperDB?
a
It seems like this really isn't v2 of the v1 Aurora Serverless, but rather "Aurora RDS with seamless autoscaling built in". For production usage (always on) this seems fantastic, but where we have 10-20 parallel envs running in dev the scale to 0 is a real blocker. Not sure it's worth paying ~500-1000/month for idling DBs. This comparison table is good: https://docs.aws.amazon.com/AmazonRDS/latest/AuroraUserGuide/aurora-serverless-v2.upgrade.html#aurora-serverless.comparison
OMG, there's an 8ACU minimum charge for RDS proxy 🤦‍♂️ If I understand correctly, that means if you want to use RDS proxy it's minimum ~$100/month even if your cluster is scaled to 0.5 ACU 😕
s
This thread totally deflated my balloon.
Sigh
Back to reading the dynamo book
t
lol
r
In theory, you don't need the RDS proxy until you're at a scale where the connection issues matter but the lack of scale to zero def hurts the most. I really wish the blog post had been clearer that v2 (as it currently stands) is more optimized towards elastic production loads and not dev/test
t
yeah this is clearly an improvement for normal users of RDS, it's not meant as a product geared to people doing full stack serverless
l
Then why did they call it serverless V2 😂
r
Because as we've seen with other launches, Serverless is either just a buzzword or "you don't specify instance types and manual scaling" to AWS. They're not on board with most peoples assumptions of the pricing floor being zero for little to no use (scale to zero).
What's funny is I sorta get why v2 doesn't scale to zero. The docs imply it's because then the cache gets nuked and there's a rehydration that happens on scale up. But I feel that's just a thing that should be documented and people understanding that can opt into
t
yeah scale to zero was basically opt in for v1. It didn't make sense for production use but for dev envs it made sense
r
Here's hoping it (and for those who use it, Data API) get re-introduced at some point soon.
g
Bumping this thread again. I'm concerned at how AWS will be treating v1 in the future (whether they'll deprecate it), how many of you are using it in production environments?
t
AWS rarely deprecates anything so I'm not immediately worried about this. Pretty sure they still have documentdb going. This is of course just a guess but it feels to me like AWS is taking 2 different approaches with RDS Serverless v1 and v2 because there's requirements that fit one or the other. Especially with services like Planetscale and Neon, think there's a lot of pressure here. But also maybe use dynamo 😄
g
I see thanks for confirming 👍 !
3 Views
Open in Slack
PreviousNext
Powered by