fetch failed during build process
Hello, this is the build id
5d2b77a7-9b6c-4565-b2cb-6372c8bab494
During the build process are fetched some files that are stored into a R2 storage named mostrum-cms
As you can see from the build logs there is this error 12:37:00.875 [error] Can't load image https://*****_454af3e47a.jpg: fetch failed
This error doesn't occur every time or on the same image resource, could you pleas check from your network logs what's the problem? Thanks55 Replies
@Cloudflare (Server Owner) hello, can someone help me? Thanks
UP
hello @Community Champion sorry for the direct tag, could you please check my request? I'm still stuck. Thanks
Don't ping champs for support
Please don't ping roles or community members for support. By pinging people they're actually less likely to help you since you've interrupted whatever they were doing or in the middle of.
I understand but I don't get support and my issue still persists
I'm a CF pages paying user.. what should I do?
This community is not an official support channel, to get official support you need to create a support ticket. Pinging people in the community is not the way to go, they are here to volunteer in their own time and not to provide professional support.
That includes myself, though I work for Customer Support my presence here is of my own volition and I am volunteering time to help out here
I understand, but I can't open tickets.. Workers&Pages paid plan don't let to create tickets.
Ok, I don't have a business CF account but I only need support on the service I'm paying for.
So please, could someone help me what is the problem?
What are you using to fetch the images? Is there a way to get more verbose errors/ retry with it
Hi @Cyb3r-Jok3 , I'm using an external library that renders a vuejs component and takes a screenshot to create a png resource. The vuejs component inject another image resource linking it using a CF R2 storage public URL stored on the same CF account.
What is strange is that with the same resource the 1st run get the error, if I start a new build it doesn't get the error.
Should be anything related to the R2 DNS or any kind of network routing?
Without better logs can’t tell the reason it would be failing. I would check to see if you can increase the logging level of the external library
@Cyb3r-Jok3 I can't increase the logs cause the web app is built by CF directly, building it locally with wrangler works fine. Please look at the build log
5d2b77a7-9b6c-4565-b2cb-6372c8bab494
I’m not a Cloudflare employee so I don’t have access to any logs. You can try using node 20 as I know there were some issues with node versions under with how the TLS connections were handled.
@Cyb3r-Jok3 ok sorry, didn't know. How can I get support from CF technical team? Thanks
?support
To contact Cloudflare Support about an issue, please visit the Support Portal and fill in the form on the portal. After submission, you will receive confirmation over email.
Some issues, such as Account or Billing related issues, cannot be solved by the community. Any plan level can open a ticket for these topics.
For more information on the methods by which you can contact Support for your plan level, see Contacting Cloudflare Support - Cloudflare Docs
Hello, I can't open tickets, I'm only paying for pages and workers pro plan. At that link CF suggest to ask here on Discord or in the community forum (already done both).
Is there anyone from the tech team here?
Yes we have but you should open ticket in support portal after CF outage solves.
@Minsu Kim I can't open tickets, workers&pages paid plan doesn't allow it. 😩
CF team help 😞 yesterday I got the error again... please tell me why the http request resolution fails.
Is this the URL? It returns a 404…
https://cms-bucket.mostrum.com//1/app_design_progettaztone_454af3e47a.jpg
I think there's a typo in your URL (for Matteo) - https://cms-bucket.mostrum.com//1/app_design_progettazione_454af3e47a.jpg
Oh yeah, there is a t instead of an i 🙈
It doesn’t load here, too…
It is just infinitely spinning in Discord for me though
Opens fine in a 
Some security thing? I assume it’s on CF, @Alessandro, check the firewall logs 🙂
Hi! ok, let me check..
I doubt CF keeps full unsampled logs of every network request from every build anyway, the environment you get is a pretty full featured Ubuntu os though, you could just change your build command to curl that file for example
curl https://cms-bucket.mostrum.com//1/app_design_progettazione_454af3e47a.jpg -vvv
i did it to test and it looks ok, no challenge for mine, worth checking though, could be a different User Agent the build tool uses or something, or the image really is messed up in some way
What is strange is that with the same resource the 1st run get the error, if I start a new build it doesn't get the error.Sounds like cache may be transforming it or something? hmm, good question would be if you get the same issues locally
Looking at sec logs I didn't find errors, @Chaika locally I don't get any issue locally and the build cache is disabled
no but the actual image itself is cachable/getting cached after the first request, and CF can transform images in cache (ex. that's the place where polish/etc gets applied, if enabled)
(Not by the Pages project, but by virtue of the image being behind cf cdn)
Looking at the DNS config, the "cms" CNAME rule is proxied by CF. Maybe is it the problem? 🤔 I can try to disable it
But if it hit the firewall it should be there…
If you're talking about my mention of unsampled logs, he was asking for the start for the Pages team to look at the network logs from the build, my comment was about that
If you can, it's not a bad idea, should give more info one way or another, would also exclude any possibility of the request being blocked
Is it a bad practice to keep CF proxy enabled on a DNS R2 rule?
oh lol it's an r2 origin? Then proxied or not it's always going to be through the cdn
thought you couldn't even edit those anyway
Yes, it's a R2 resource
Maybe accessing it through the auto generated R2 public url? (so without using the personalized one)
Do you have any other images being loaded from that bucket/custom domain?
yep, a lot, only one per build fails.. randomly
Randomly in the sense it's a different one each time, or it's the same one, failing randomly?
different one each time
Any chance you can just use the internal binding to fetch those instead of doing the normal fetch?
but if i start another built in the closest to the prev one all images are loaded correctly
it's during build, not function/worker execution
yep, I can try
Oh, dumb me 🙈
would be nice though
the r2 pub domain is ratelimited and not meant for production
DId you get a chance to check Security -> Events as matteo suggested?
Sure, already checked, here it is:
yesterday (6th nov) I got the last build issue
as you can see the activity log is empty
What's the build setup you're using, nitro with some framework? Any way to get better logs out of it then "fetch failed"? Increase timeouts, etc?
The web app is made using vue/nuxt, and it's build using the nitro preset config.
I have already asked the owner of the js lib that is causing this issue if it can be related to a timeout but increasing it didn't resolve the problem.
Essentially, this lib create a png file using satori lib, satori make a screenshot of a vue component when rendered. In the vue component there is an html tag that uses the failing image resource as a background css rule.
hmm, so do you know where that error is being thrown from? I can't find
can't load image or fetch failed in satori or nitro
Could it be nuxt server-side rendering that, and satori throwing that and it's reflected down the build? Weird too that it doesn't stop/kill the build, wonder if you couldn't put in some retries there or somethingLooking at satori source code I can find the error
can't load image but searching with github in the current last version it doesn't exist
I must try to update my dependency (if I can cause in this case satori is not a primary dependency)
Let me investigate, I'll come backInteresting.. I see it was removed here: https://github.com/vercel/satori/commit/52fa760e63656900f482e85a5ffc481dd2b4b818 and some more image types were supported
When it errors out like that, does it actually mess up the output (i.e the image is missing)?
actually nvm even in the latest version it's still there at the same spot: https://github.com/vercel/satori/blob/a2abe37948071414ba93d2e084e1c09dad5c4899/src/handler/image.ts#L266C13-L266C13
Still, worth updating and seeing if they haven't fixed anything
do you intend the final rendered image?
My question was just if it effects it at all yea, it looks like there's a few other moving parts there with cache and whatnot
simply satori renders it as a black image
yep, satori uses something like a cache layer but if I'm not wrong I have disabled it
(if I remember well)
I don't think you did considering the image file has this at the top
* TODO: We might want another option to disable image caching by default.lol
omg lol ahah, in this case the cache I have disabled works for something different
maybe it cache the rendering result
(i'm rebuilding the app with the last satori version)
ok, nitro builded without errors, but I can't be sure the issue is resolve, this could be one of the cases where it works 😦
I'll monitor for now, thank you guys for the hints, really appreciated