serverless functions tree-shaking
Unanswered
American Bobtail posted this in #help-forum
American BobtailOP
Hi 👋
i was wondering why each serverless function deployed on vercel would have the same size, even though
as i understand it, each handler uses the same global
(in my screenshot example, i import the
is there a reason for that or could we think of a way to change the nextjs build script to include a specific
i was wondering why each serverless function deployed on vercel would have the same size, even though
@vercel/nft gives indications of the deps of each handler.as i understand it, each handler uses the same global
package.json which can be quite heavy and take a toll on lambda cold starts.(in my screenshot example, i import the
sharp library in only one of my 2 handlers, yet both of them end up with a 15.5MB size.)is there a reason for that or could we think of a way to change the nextjs build script to include a specific
package.json for each handler with only the dependencies it needs itself?
1 Reply
American BobtailOP
bump + a little more context:
over the weekend i looked at the source code of
so my new question is, shouldn't the functions size ignore the layer size ?
let's say i have a dependency A that weighs 30MB, and dependency B that also weighs 30MB.
if i try to deploy handler1 with dep A and handler2 with dep B i'll end up with a layer of 60MB (aws limit is 250MB for a layer) but vercel will tell me that each of my functions weight 60MB (aws limit is 50MB for a lambda)
i'm guessing vercel won't even try to uplaod to aws since it'll detect 60MB > 50MB even though each of my function should only weigh 30MB.
over the weekend i looked at the source code of
nextjs and vercel/next builder and ended up understanding that the lambas are grouped, and a layer is created with all the dependencies, so that any handler has access to anything it might need.so my new question is, shouldn't the functions size ignore the layer size ?
let's say i have a dependency A that weighs 30MB, and dependency B that also weighs 30MB.
if i try to deploy handler1 with dep A and handler2 with dep B i'll end up with a layer of 60MB (aws limit is 250MB for a layer) but vercel will tell me that each of my functions weight 60MB (aws limit is 50MB for a lambda)
i'm guessing vercel won't even try to uplaod to aws since it'll detect 60MB > 50MB even though each of my function should only weigh 30MB.