High CPU usage when running @trigger.dev/cli@latest | Trigger.dev | Page 1

prisma dagger Jan 12, 2024, 11:45 AM

#

I am using NextJS, when I run npx --yes @trigger.dev/cli@latest dev node server CPU usage goes above 200%. Not sure how to debug it...

haughty karma Jan 12, 2024, 11:50 AM

#

I can think of two things that could be causing this:

You're doing a lot of runs?
There's a bug in our new tunnel (yalt) which is causing high CPU usage.

Are you doing a lot of runs? If we can rule that one out then it's probably something on our end 🤦‍♂️

blazing parrot Jan 12, 2024, 12:48 PM

#

I'd be very surprised if it was the yalt.dev change... that simply opens a WebSocket connection (using ws), but of course there could be something going wrong there that's causing an issue

prisma dagger Jan 12, 2024, 1:46 PM

#

I don't get to do any runs - as soon as I run Trigger CLI CPU usage goes up, and I can't really use my Mac until I kill the process

#

process is next-server not the Trigger CLI itself

haughty karma Jan 12, 2024, 1:58 PM

#

Hmm interesting. When you just run your Next server, without our CLI, it's all fine?

prisma dagger Jan 12, 2024, 3:15 PM

#

yes

haughty karma Jan 12, 2024, 4:34 PM

#

Sorry for the slow response

#

Do you have a lot of runs that are queued on dev? So when you connect it suddenly starts a ton of runs? You can monitor this in the new "Runs" page and filter by the "Dev" environment.
Do you have a lot of Jobs? I'm wondering if the indexing is sending a ton of JSON, but that seems unlikely.

prisma dagger Jan 13, 2024, 4:49 PM

#

I have no queued runs
I have 9 jobs

haughty karma Jan 15, 2024, 11:23 AM

#

We're trying to reproduce this, we've not seen it before.

Could you send us the output of this: npx envinfo --system --binaries and your version of Next.js?

prisma dagger Jan 15, 2024, 6:59 PM

#

I in San Francisco until 21st, I will send those when I get back

prisma dagger Jan 23, 2024, 7:39 AM

#

System:
OS: macOS 12.7.2
CPU: (8) x64 Intel(R) Xeon(R) CPU E5-1620 v2 @ 3.70GHz
Memory: 13.35 GB / 32.00 GB
Shell: 3.2.57 - /bin/bash
Binaries:
Node: 18.19.0 - /usr/local/bin/node
Yarn: 1.22.4 - ~/.yarn/bin/yarn
npm: 10.2.3 - /usr/local/bin/npm
Watchman: 2023.02.27.00 - /usr/local/bin/watchman

#

"next": "^14.0.3"

haughty karma Jan 24, 2024, 10:10 AM

#

Going to try and reproduce this by running the same versions of Node and Next

prisma dagger Jan 30, 2024, 11:28 AM

#

I tried updating trigger to newest version, but no change

#

It looks like nextjs Compiling /api/trigger ... never finishes - like it is stuck in infinite loop

#

updated to nextjs "14.1.0", no change

#

I removed all my jobs from index file and just added the example job

#

$ next dev
   ▲ Next.js 14.1.0
   - Local:        http://localhost:3000
   - Environments: .env.local, .env

 ✓ Ready in 5.8s
 ○ Compiling /api/trigger ...
 ✓ Compiled /api/trigger in 22.6s (41 modules)

#

it did compile, but CPU usage does not go down

#

I am bearly able to use my computer, until I kill the process, so it is difficult to do debugging

haughty karma Jan 30, 2024, 5:12 PM

#

I'm trying to reproduce this now

#

Are you using the app router?

#

I just tried to reproduce with the create next-app settings

#

Next.js 14.1.0
With the example job.

#

And I get normal seeming compile times of

#

I think we need to try figure out if it's something special about your project.

Can you try doing this, it should only take 5 mins:

Create a new project in Trigger.dev
Select Next.js from the frameworks grid
Select "Create a new Next.js project"

There are 5 steps on there.

prisma dagger Jan 30, 2024, 9:28 PM

#

haughty karma Are you using the app router?

yes

prisma dagger Jan 30, 2024, 9:29 PM

#

haughty karma I think we need to try figure out if it's something special about your project. ...

done - works without any problems

haughty karma Jan 31, 2024, 9:18 AM

#

OK so it's something about your project. What's your dev script in your package.json?

#

I know some people have had luck using next dev --experimental-turbo

prisma dagger Jan 31, 2024, 10:04 AM

#

I guess it is just --turbo now. It did solve the CPU issue! But now I have a new problem:

#

yarn run v1.22.4
$ next dev --turbo
   ▲ Next.js 14.0.4 (turbo)
   - Local:        http://localhost:3000
   - Environments: .env.local, .env

 ✓ Ready in 1982ms
 ○ Compiling /api/trigger ...
 ✓ Compiled /api/trigger in 15.4s
 ○ Compiling /_error ...
 ✓ Compiled /_error in 1151ms

#

$ npx --yes @trigger.dev/cli@latest dev
✔️ [trigger.dev] Detected TriggerClient id: segment-x--jZ_
✔️ [trigger.dev] Found API Key in .env.local file
✖ [trigger.dev] Server responded with 500 (http://localhost:3000/api/trigger).
✖ [trigger.dev] No server found (http://localhost:3001/api/trigger).
✖ [trigger.dev] No server found (http://localhost:3002/api/trigger).
✖ [trigger.dev] Your endpoint couldn't be verified. Make sure your app is running and try again. /api/trigger
  [trigger.dev] You can use -H to specify a hostname, or -p to specify a port, or -s to specify https, or -t to specify the tunnel-url pointing to the local dev server.
✨  Done in 25.25s.

haughty karma Jan 31, 2024, 10:05 AM

#

It's crazy that it takes 15 seconds to compile /api/trigger still

prisma dagger Jan 31, 2024, 10:05 AM

#

I tried adding verbose to trigger client, but that does not get me the error message

#

btw. when I push code to staging (Vercel) enviroment, everything works fine

haughty karma Jan 31, 2024, 10:07 AM

#

Yeah the Next.js dev server since 13 is terrible

#

Although no one else has had performance issues like this

#

You're on a newish version of the @trigger.dev/sdk?

prisma dagger Jan 31, 2024, 10:10 AM

#

2.2.11

haughty karma Jan 31, 2024, 10:10 AM

#

I can't see us returning a 500 from anywhere. That failure that's happening is our CLI trying to connect with http://localhost:3000/api/trigger. So it doesn't go via the internet.

#

Ok try updating to see if that helps: npx @trigger.dev/cli@latest update will update all your trigger packages

#

I'm wondering if they've changed to ip v6 with that command. Astro does that.

#

If updating the packages doesn't work try re-running the CLI:

npx --yes @trigger.dev/cli@latest dev -H "[::]"

That's the IP v6 localhost

prisma dagger Jan 31, 2024, 10:19 AM

#

I don't think that's it, /api/trigger is found and compiled, but throws a 500 error, any idea how can log it?

#

btw. I updated to 2.3.17

haughty karma Jan 31, 2024, 10:23 AM

#

Yeah good point, it wouldn't get a response at all

#

You can add:

verbose: false,
logLevel: "info",
ioLogLocalEnabled: true,

to the new TriggerClient() options.

#

But we might not be logging this out. I've never seen a 500 come back from the /api/trigger endpoint before

prisma dagger Jan 31, 2024, 10:31 AM

#

so that didn't give me anything back, but when I have my app open during /api/trigger compilation I got that:

haughty karma Jan 31, 2024, 10:34 AM

#

mjml isn't a package we use, I just searched our pnpm-lock file that all our dependencies and their dependencies

#

I think this --turbo flag is probably experimental for a reason

prisma dagger Jan 31, 2024, 10:36 AM

#

so - removing email sending task from trigger jobs solved the issue

haughty karma Jan 31, 2024, 10:36 AM

#

What are you using to send emails?

prisma dagger Jan 31, 2024, 10:37 AM

#

resend + react-mjml parsing function that is the issue, I guess

#

In the past I added

  experimental: {
    serverComponentsExternalPackages: ['mjml']
  },

to my next.config.js

haughty karma Jan 31, 2024, 10:39 AM

#

If you add mjml-core as well does it fix it?

#

I've had to do stuff like this in Remix before, it's a pain

#

I had to add a load of packages to it, even dependencies of dependencies

#

I've not heard of react-mjml before. Are you using it because they have more components?

prisma dagger Jan 31, 2024, 10:42 AM

#

haughty karma I've not heard of `react-mjml` before. Are you using it because they have more c...

I tried using resend's react-email in beta, but it was still not production ready, a lot of styling/responsive issues. I used pure mjml for my last project, but I liked idea of having the code in react so react-mjml was perfect 😉

#

I added mjml-core to serverComponentsExternalPackages but now I just get different error related to it

haughty karma Jan 31, 2024, 10:47 AM

#

This looks like an issue with dynamic imports. Turbo probably does it differently. I had a similar issue with esbuild recently that was fun to fix.

#

The docs are a bit thin on htis: https://nextjs.org/docs/pages/api-reference/next-config-js/turbo

next.config.js Options: turbo | Next.js

Configure Next.js with Turbopack-specific options

#

If you're using resolve.alias is looks like you'll have to change some things

prisma dagger Jan 31, 2024, 11:06 AM

#

looks like fun

prisma dagger Jan 31, 2024, 11:42 AM

#

ok - I think I am ok - back to next dev without --turbo, with added mjml-core in serverComponentsExternalPackages and updated Trigger

haughty karma Jan 31, 2024, 11:42 AM

#

It's behaving itself?

#

I have had slowness before with Node – removing all the node_modules folders from the monorepo and then doing a fresh install of packages solved it.

prisma dagger Jan 31, 2024, 11:45 AM

#

haughty karma It's behaving itself?

I think so, so far so good

haughty karma Jan 31, 2024, 11:45 AM

#

Ok great

prisma dagger Jan 31, 2024, 11:48 AM

#

Thanks for helping me debug this issue!

haughty karma Jan 31, 2024, 11:49 AM

#

No problem prayge

prisma dagger Feb 1, 2024, 11:48 AM

#

So, I am back where I started, I needed to downgrade Trigger to 2.2.11 because of that issue: #1186360299952029717 message
Now I again struggle with CPU usage skyrocketing when trying to use Trigger in dev

haughty karma Feb 1, 2024, 2:41 PM

#

So you were getting those build errors again?

#

I saw this similar problem and the webpack changes required to fix it: https://stackoverflow.com/a/77738096

Stack Overflow

Webpack ERROR in node:crypto Module build failed: UnhandledSchemeEr...

I am attempting to update some older code to eliminate warnings. I have upgraded some code that uses the 'crypto' module. However, webpack 5.88.2 does not recognize the new syntax. My new import lo...

#

I haven't seen this issue with any other customers, I think it might be a Next.js issue with pages + app routers maybe

prisma dagger Feb 9, 2024, 4:17 PM

#

I solved the other issue, now I am on Trigger.dev 2.3.17, but high cpu usage was still occuring, just after successful trigger route compilation

#

finallyI tried setting up my ngrok tunel - and this seams to have solved the issue

#

Is there a way to automaticly create ngrok tunells with latest version similar to behavior of npx @trigger.dev/[email protected] dev ?

haughty karma Feb 9, 2024, 4:26 PM

#

That's interesting that its being caused by our tunnel. No there isn't but you can just run that old version of the CLI and it should work fine

prisma dagger Feb 9, 2024, 4:48 PM

#

fun fact: npx @trigger.dev/[email protected] dev works for me but not for my college on Linux, but he have no problems with @latest 😜

haughty karma Feb 9, 2024, 4:55 PM

#

Ha ok, probably we made some fixes that were impacting some users

#High CPU usage when running @trigger.dev/cli@latest