"Payload has transformed the way our clients manage content. It's an indispensable tool for any modern agency."
Get the resources you need to start building today
Microsoft chose Payload to tell the world about AI.
/admin/login doesnt function as expected
Hey all, were facing some issues with deploying the payload-next-demo app. We only face this when deploying with SST.
Essentially, the /api/users/login is route is not erroring (no log output), but we receive a 200 ok response with no data. We are unable to reproduce this locally, but wondering if anyone who is slightly more familiar with the codebase can help. I have also tried deploying a basic next app and attaching payload-next myself, which yields the same results.
We see no client side failure, but nothing actually happens, you hit login, which is followed by a 200 with no data, and you are not logged in. We do not see any issues at other API routes, like /api/users/me, or /api/users/init.
Have dug some more into this. We are seeing a status 200 returned, with no error message, and no payload. I have attempted invoking /forgot-password instead, and there we see a 500, which, has a key difference of returning our stack and an error message, rather than the unexpected 200 without data.
narrowed it down some more, on successful post requests, we see no data coming back to us
futher narrowed down, the 'body' is not making it to the lambda, nor is the expected data under any of the expected fields - this is fixed with bodyparser set to fasle
It appears
'payload/dist/auth/operations/login'is returning with res.status somewhere, with a 200. Any help is appreciated
Added a beforeOp and beforeLogin Hook to trace. beforeOp logs are successfull, but no beforeLogin hook gets run. Narrows down the issue to this section (
https://github.com/payloadcms/payload/blob/master/src/auth/operations/login.ts#L56-L130)
Seems like, from a walkthrough it can be narrowed down to either: lines 96-98 or 126
Seems like we can rule out line 126, as it seems we should have guarenteed ourselves an auth error before that point on line 102 (purposely using wrong creds) - seems reasonable to assume its an issue somewhere in the auth function
one possible spot: this should probably use Buffer and not the string of password if im not mistaken?
https://github.com/payloadcms/payload/blob/master/src/auth/strategies/local/authenticate.ts#L23- not sure this is our root cause though
possibly narrowed down more to an issue where lambda isnt allowing a proper connection to mongo - seems like it works on startup but is somehow intermittently failing with findOne, but no logging
strange, simply calling
console.log(await Model.findOne({ email }));is causing this issue it seems. When this is executed in the hook, the lambda instantley ends with a 200
bumping this as were not seeing anything coming through still, could this be to do with the config of the mongoose client?
@844209275089453076When you say that you click login, do you mean from the Payload admin panel?
Where does the login "failure" occur
Yea
The login doesn’t actually “fail” the request is returned with 200, but with no response
(Which, is basically a no op on the admin panel and just stalls)
Notable, in the lambda logs we see no error message either, we added some custom logging and saw that it seems to end directly after the FindOne
Repro steps are: clone the demo, yarn sst create and then yarn sst deploy
- @858693520012476436
This person has been stuck for a bit, any ideas on this? You've worked with serverless before, right?
Actually, I think paul only has deployed on Vercel
@281120856527077378Any ideas on this one? 😄
@844209275089453076On a side not while we are waiting for some folks with more knowledge
Have you tried hitting the API instead of logging in, as a test?
Via like postman, etc
Yes
Yeah I only deployed on vercel and migrated that project away from serverless in the end actually
No difference
Aws > vercel sorry not sorry
I’ve isolated this all the way down using a before op hook, and running a direct Model.findOne on a hardcoded email
same thing under the hood...the whole internet is 😭
IMO REST + selfhost is best
only because i never run into any complicated issues hehe
@844209275089453076I feel like the findOne discovery is a symptom of the issue but not the cause
But to get no response data back is super weird
especially if the requests come back 200
@844209275089453076Do you know if there were any oddities when configuring your payload server settings, etc
CORS, CSRF, anything odd?
Also, if you add a standalone route in server.ts, just a base express route, can you hit that?
If not then I think that is telling
This is from the lambda to be clear
It kills the lambda somehow
It’s incredibly odd
I’ve engaged mongoose and I’ll try engaging aws support
💪
And the Lambda doesn't throw any errors / log?
No lol
I’ve basically got uh
console.log(“blah”)
Model.findOne({email:”test@test.com”});
Console.log(“I got here”)
And I get
blah
END REQUEST UUID
Sorry for the horrific formatting, let me like the mongoose discussion incase it’s from there
https://github.com/Automattic/mongoose/discussions/13852To be clear, I’m not clear on if this was a payload issue but was looking for help it’s sipoer weird
Is the mongodb self hosted or through atlas?
Even inside a try catch btw
Atlas serverless
Another point tho- it’s able to create the collections
Oh it doesnt hit the catch?
@844209275089453076Can we check out how your payload config is setup?
The user has full admin perms
Also, had it wrapped in try catch and saw nothing!
On another note the lack of logging has been a bit frustrating tbh- I’ll have to get approval from my company before working on anything FOSS but once I do I’ll look at raising some PRs
Yh- but am at work for time being
Exactly
ok so, i know this isnt great
MONGODB_URI=mongodb+srv://USERNAME:PASSWORD@serverlessinstance2.ozlmxxx.mongodb.net/?retryWrites=true&w=majority PAYLOAD_SECRET=YOUR_SECRET_HERE PAYLOAD_CONFIG_PATH=dist/payload.config.js NEXT_PUBLIC_APP_URL=https://xxxx.cloudfront.net PAYLOAD_PUBLIC_CMS_URL=https://xxxx.cloudfront.net S3_ACCESS_KEY_ID= S3_SECRET_ACCESS_KEY= S3_REGION= NEXT_PUBLIC_S3_ENDPOINT= NEXT_PUBLIC_S3_BUCKET=but this
I gave PayloadCMS 2.0 w/ Postgres using NextJS on SST a shot. Unfortunately the same error occurs even with Postgres. I was able to create the first user, but I can't Login.
Super weird, is that still using mongoose somewhrrr then?
No, Mongoose is isolated to the db-mongodb adapter now. There isn't anything in core using it.
All this being said, we have not attempted to look into what is needed for SST support.
It has nothing to do with the database of choice. It seems like a re-direction does not happen after the login/first-registration
I can't say we've put any effort into supporting SST at this point. Pretty niche.
It's niche until you migrate to nextjs. Vercel is great, but not always the solution from a legal point of view.
hey guys, I'm having the same problem with PayloadCMS 2.0 (payload-next monorepo setup) & SST. I've tried postgres (supabase) and mongodb (atlas) and in both cases I wasn't able to log in, while recieveing reponse code 200 :/
@838356081943773204, just wondering did you try postgres using SST RDS serverless cluster?
Able to share what is the legal point of view for using Vercel?
This is the exact same setup I POC-ed for as well. Same result
Sorry for the late reply. Legal as in being GDPR compliant and making really sure, that the PII data is not send across borders.
Oh I see
Thanks for the explanation
Star
Discord
online
Get dedicated engineering support directly from the Payload team.