Does API function timeout if it’s triggered by multiple cron jobs?

I need to send a newsletter to 20k subscribers. Resend doesn’t yet support bulk email sending, so I want to build my own using Resend, Vercel Cron and Next.js API route.

I am not certain though, will a serverless function timeout if I keep triggering it via cron every second for like half an hour?

Neto•8/8/23, 7:41 PM

each function call spawns a new lambda

edgarasOP•8/8/23, 7:41 PM

Myeah. But there is something about warm/cold lambda

Neto•8/8/23, 7:42 PM

cold lambda is when the lambda takes some time to start

Neto•8/8/23, 7:42 PM

with 20k calls instantly

Neto•8/8/23, 7:42 PM

some of them will suffer from that

Neto•8/8/23, 7:42 PM

majority of them wont

edgarasOP•8/8/23, 7:42 PM

Yeah but once it’s warm, wouldn’t it try using the same warm lambda? And then eventually timeout?

Neto•8/8/23, 7:42 PM

you should 100% do a blank run with a smaller amount and see how it handles

edgarasOP•8/8/23, 7:43 PM

Definitely. I am still not sure this is the best way to do this in serverless

Neto•8/8/23, 7:43 PM

that aspect of routing between lambdas is only to aws to deal with

Neto•8/8/23, 7:43 PM

in the real world

Neto•8/8/23, 7:43 PM

if you batch those functions calls

Neto•8/8/23, 7:43 PM

like 500 blocks each

Neto•8/8/23, 7:44 PM

most of them will hit the same warm lambdas after some calls

edgarasOP•8/8/23, 7:44 PM

I will try to batch a bit also. Resend can send 10emails a second, so I can send around 200 emails in 30 sec in a serverless function

Neto•8/8/23, 7:44 PM

about the timeout

Neto•8/8/23, 7:44 PM

iirc the pro version of vercel is around 60s

edgarasOP•8/8/23, 7:45 PM

Yes but it says better run 30-60 secs. And generally lambdas are made to run up to a few secs max

edgarasOP•8/8/23, 7:46 PM

The reason I am asking because I was confused about a warm lambda, if it will try to use the same one for same job cuz it’s warm, and eventually timeout. Need to test that. On localhost the response time was adding up for some reason, the more I was calling same function. But I was testing it recursively, maybe that’s why

Neto•8/8/23, 7:47 PM

you can check this video to have a different idea on how to do it

Neto•8/8/23, 7:47 PM

https://www.youtube.com/watch?v=xgUpI_EnAQs

YouTubeWeb Dev Cody

How I generated 10,000 PDFs in seconds using SST

T3 Stack Tutorial: https://1017897100294.gumroad.com/l/jipjfm

SaaS I'm Building: https://www.icongeneratorai.com/

Discord: https://discord.gg/4kGbBaa

Newsletter: https://newsletter.webdevcody.com/

GitHub: https://github.com/webdevcody

Twitch: https://www.twitch.tv/webdevcody

Website: https://webdevcody.com

Twitter: https://twitt...

Neto•8/8/23, 7:47 PM

main issue probably will be the first lambda, the trigger, to be alive for the 20k

Neto•8/8/23, 7:48 PM

you can probably do a "fan out" approach to it

edgarasOP•8/8/23, 7:50 PM

Thnx I will try looking into it! Long running jobs in Vercel/Next are not very well documented…

Neto•8/8/23, 7:51 PM

Christos•8/8/23, 7:51 PM

To be fair lambdas / serverless function aren't designed for long running jobs I'd say

Neto•8/8/23, 7:52 PM

with aws you have more flexibility around lambdas

Christos•8/8/23, 7:52 PM

You can always rent a VPS ( a server that you own ) and run your job there

Neto•8/8/23, 7:52 PM

with vercel you are tied to that nextjs should/can do

edgarasOP•8/8/23, 7:54 PM

Yeah… I know. I would just like to keep it all in one code base. Just to send this newsletter now I have to start VPS, maintain it etc.). I rather have something that is hacky and works but in the same Next.js codebase.

edgarasOP•8/8/23, 7:55 PM

So Vercel Cron + multiple serverless function triggers was my last hope before getting to something like Qstash, or running a separate node.js app in VPS…

NNeto Click to see attachment

edgarasOP•8/8/23, 7:57 PM

Here you call the functions without requiring to return a response right?

Neto•8/8/23, 7:57 PM

yep

Neto•8/8/23, 7:58 PM

if they fail for some reason, sucks to suck

edgarasOP•8/8/23, 8:01 PM

Well, I will be calling my Postgres db on every call, checking for unsent emails and then after successfully sending, marking every user row as sent. So should be fine even if some of them fails. Failed ones will be retried on other calls.