Appwrite Sites - getaddrinfo ENOTFOUND

Has anyone had a problem having appwrite on a subdomain in the project custom domains? We have an issue with the Appwrite Sites. Sometimes the internal node:dns (runtime node-22) has a problem getting addrinfo, luckily we use Sentry so we know this type of error only happens 3-6 times a day. Should we create a PR?

We are using the subdomain as endpoint, because it is recommended by the docs (Appwrite is using localStorage for session management. Increase your security by adding a custom domain as your API endpoint.) , so instead of https://fra.cloud.appwrite.io/v1 we use https://appwrite.rostlino.cz/v1

https://discord.com/channels/564160730845151244/564160731327758347/1398253413841633321

can you try and output more details such as the stack from the error?

can you share the exact timestamp it happened last?

@Steven Hi Steven, sending you a public sentry link https://rostlino.sentry.io/share/issue/d4a5a1f4bfff421ebca39777aed0c0d6/

As you can see, the issue is at the server side of Appwrite at node:dns

Error: getaddrinfo ENOTFOUND appwrite.rostlino.sk
  File "node:dns", line 120, col 26, in GetAddrInfoReqWrap.onlookupall [as oncomplete]
  File "node:internal/async_hooks", line 130, col 17, in GetAddrInfoReqWrap.callbackTrampoline
TypeError: fetch failed
  File "node:internal/deps/undici/undici", line 13185, col 13
  File "node:internal/process/task_queues", line 105, col 5, in process.processTicksAndRejections`

Currently looks like, that it is happening only when using custom subdomains, and it looks like **global ** issue that might affect **functions ** as well when using the endpoint as custom subdomain.

After the expiration of 3600sec when the DNS is again resolved (proximately 2sec), the domain works (looks like for another 3600sec), so probably some issues at Fastly? Cold-boots?

Can you try adding a DNS record for the domain and point it to the region endpoint?

@Steven Hi Steven, what do you mean adding a DNS record for the domain? Do you mean subdomain? CNAME is wildcarded so the subdomain is included and is pointing to appwrite.network by default.

Do you mean like include the CNAME that points to the API endpoint?

yes, like that

Configured, currently waiting for the DNS's to update, we have multiple domains, so let's focus only on the rostlino.com
https://dnschecker.org/#CNAME/appwrite.rostlino.com

@Steven Also, we have disabled the logging (few days ago) on all the sites and yet, still getting the logs, It will be another issue, so I will create another ticket.

@Steven The issue is still even after updating the DNS record CNAME to appwrite, so the appwrite.network is not used to resolve the appwrite subdomain.

Is it happening less?

@Steven Looks like it's fixed 👍

FYI, we pushed another improvement. Is it still going okay for you?

@Steven Hello 👋 , adding CNAME record to the fra.cloud.appwrite.io resolve the issue.

I will remove the CNAME record, on one of the domains (rostlino.com) so it will fallback to the wildcard with value appwrite.network and let you know, if the strict CNAME record is required or not.

@Steven Hello 👋 , looks like the CNAME record is no longer needed and the wildcard for appwrite.network works correctly now 🙌

@Steven Looks like the issue is still present.

the issue went away and came back?

how frequent does it happen?

Yes, now it's with every request, so all the time.

@Steven Any updates?

what if you replace that fra.cloud.appwrite.io cname with dualstack.t.sni.global.fastly.net?

Will try and let you know
https://dnschecker.org/#CNAME/appwrite.rostlino.cz

@Steven The issue is still present