Slow API fetch, both in dev and prod

[crnls]

The last few days I have been trying to fix extremely slow API fetch issues with my Medusa store. I have tried many obvious solutions, but so far without success. Screenshots below with logs, there are no console errors. Add to cart often takes +6 seconds, simple product pages 2-4 seconds.

Notes:

• Locally, with external DB, API is also slow
• Admin API is fast
• Backend at Heroku (https://dev.wantsandneeds.nl/)
• Admin & Storefront hosted at Vercel (Vercel functions now build at Frankfurt server) (https://admin.wantsandneeds.nl/ & https://wantsandneeds.nl/)
• Database at Supabase (Frankfurt, closest to traffic (Netherlands))
• Tested everything also in clean install, also slow fetch there.
• Tested several DB providers (AWS RDS, multiple Supabase servers etc)
• Cart interactions are especially slow

Screenshots:

[bqst]

Since the switch to the new starter pack and server actions, I've been facing the same issue. The actions, especially the 'add to cart' one, now take a significantly longer time (~2 sec) with the same backend.

[VariableVic]

Is it just as bad when you interact with the server directly? E.g.

curl -X POST '{backend_url}/store/carts/{id}/line-items' \
-H 'Content-Type: application/json' \
--data-raw '{
    "variant_id": "{variant_id}",
    "quantity": 1
}'

I'm running the same setup for next.medusajs.com. The only difference is that I host Redis and Postgres on Heroku instead of Supabase. So that might be the bottleneck here.

Still weird, since you tried several DB providers. Really hard to tell what's causing this bad performance.

[crnls]

Hey Victor,

Redis is for me also at Heroku, db Supabase.

This shouldn’t be an issue right?

When I interact with the server directly, the load is about 2 seconds.

Groet, Martijn van der Pas

Op 14 mrt 2024 om 12:31 heeft Victor Gerbrands @.***> het volgende geschreven:



Is it just as bad when you interact with the server directly? E.g.

curl -X POST '{backend_url}/store/carts/{id}/line-items' \ -H 'Content-Type: application/json' \ --data-raw '{ "variant_id": "{variant_id}", "quantity": 1 }'

I'm running the same setup for next.medusajs.com. The only difference is that I host Redis and Postgres on Heroku instead of Supabase. So that might be the bottleneck here.

Still weird, since you tried several DB providers. Really hard to tell what's causing this bad performance.

— Reply to this email directly, view it on GitHubhttps://github.com/medusajs/nextjs-starter-medusa/issues/280#issuecomment-1997238070, or unsubscribehttps://github.com/notifications/unsubscribe-auth/AEL5BM63ESTYSY75JNK4ZUTYYGC7ZAVCNFSM6AAAAABD2FWQP6VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTSOJXGIZTQMBXGA. You are receiving this because you authored the thread.Message ID: @.***>

[dsijakovski98]

Having the same issue, very slow responses when going through the SDK. Maybe it will be faster using regular API calls to the server directly, tho haven't tried it out yet.

Tried also running backend in cluster mode in production, didn't seem to make any difference on the performance

[crnls]

Never was able to solve this, might try at a later point if caching some stuff with vercel improves load times

Groet, Martijn van der Pas

Op 7 jun 2024 om 13:53 heeft dsijakovski98 @.***> het volgende geschreven:



Having the same issue, very slow responses when going through the SDK. Maybe it will be faster using regular API calls to the server directly, tho haven't tried it out yet

— Reply to this email directly, view it on GitHubhttps://github.com/medusajs/nextjs-starter-medusa/issues/280#issuecomment-2154680984, or unsubscribehttps://github.com/notifications/unsubscribe-auth/AEL5BM45GBYOW6NUWCCKHVLZGGNLDAVCNFSM6AAAAABI6QTRS6VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDCNJUGY4DAOJYGQ. You are receiving this because you authored the thread.Message ID: @.***>