cross-device link not permitted
I for the life of me cant get past this error.
This is what i get if i run local dev, cloud is fine.
I had gotten rid of my CONVEX_TMPDIR setting temporarily (its been on there so long i forgot why i had it)
I'm not sure its related as i was having issues with switching between various local backends from different repos.
I dont know if i had CONVEX_TMPDIR locally set somewhere. It shows I've tried new projects Added it to .env.local I deleted temp i deleted ./convex Reconfigured countless times. nvm use 18 etc etc
I dont know if i had CONVEX_TMPDIR locally set somewhere. It shows I've tried new projects Added it to .env.local I deleted temp i deleted ./convex Reconfigured countless times. nvm use 18 etc etc
14 Replies
Thanks for posting in <#1088161997662724167>.
Reminder: If you have a Convex Pro account, use the Convex Dashboard to file support tickets.
- Provide context: What are you trying to achieve, what is the end-user interaction, what are you seeing? (full error message, command output, etc.)
- Use search.convex.dev to search Docs, Stack, and Discord all at once.
- Additionally, you can post your questions in the Convex Community's <#1228095053885476985> channel to receive a response from AI.
- Avoid tagging staff unless specifically instructed.
Thank you!
When i first provision i get CONVEX_TMPDIR warning regardless it seems.
But then when i run npx convex dev later it doesn't show that warning. But then i don't get any files into my
However export CONVEX_TMPDIR=/newtemp gets rid of the error but doesnt fix it
I'm on ubuntu 24.10, seems like running convex as root would work as i get to the node 18 error but id have to downgrade node as nvm isnt setup on root.
no luck with reboot
Any ideas about what else i could try
figuring out who to ping here
cc/ @sshader
Someone here fixed a similar issue in the past by regenerating the convex folder, have you tried that by any chance?
https://discord.com/channels/1019350475847499849/1019350478817079338/1207338632260882473
ok i deleted my generated folder but for some reason what is also happening on cloud dev is i'm getting the
every time i run npx convex dev
The first time i run npx convex dev im getting
but also randomly during every dozen or two pushes
I have triple checked my CONVEX_TMPDIR but somehow it sure doesn't seem like its being read.
Before it was set to a folder that didn't exist from my old computer... now it does exist.
What do you mean by this?
However export CONVEX_TMPDIR=/newtemp gets rid of the error but doesnt fix it
I was experimenting with local env variables on ubuntu too, seeing if that would help
As i understand the CONVEX_TMPDIR must be set on the convex backend environment vars? I also have it set in .env.local.
No, you need to set it on your local machine (the one you're running
convex dev
on)
I think using .env.local will not work here
Could you try moving export CONVEX_TMPDIR=/newtmp
to your bashrc file (or equivalent for your shell)?The CONVEX_TMPDIR warning is gone but when i configure back to npx run dev --configure and select local it still creates temp files in /tmp/ nothing is getting created in my specified folder
so i still get the cross-device link not permitted
I was thinking /tmp was on my main drive but its actually in ram.
Can you please send me the whole output you're getting?
it seems to be something with this project as i can run other things locally, i just cant figure out what is different. Im trying the local deployment on different projects on convex cloud just in case.
I also notice that there is nothing in my defined TMP folder even when i use a working local deployment
Ah I see a spot where we're not respecting
CONVEX_TMPDIR
-- this is specific to running node functions that use external packages.
Sounds like you know this already, but switching to cloud should always work.
We'll work on fixing this and can report backcool. Thanks!
I'm curious if while you are at it, could it be possible to add a feature for changing up local ports quickly or just moving to new ones if that port is unavailable, i keep wanting to run a couple local deployments at a time. (I haven't had time to like dig into this, so maybe its already solved )
moving to new ones if that port is unavailableThis should already be true! Let us know if you've seen issues running multiple at a time. There's also CLI args (currently hidden) for
--local-cloud-port
(the .convex.cloud equivalent, usually 3210) + --local-site-port
(the .convex.site equivalent, usually 3211), which we can expose + make easier to use