500 error on chat completions messages GET

API Bugs
1

I’ve been getting this error 90% of the time for several weeks now since I started using this EP. For no reason I understand it’ll work for a few mins and then stop again, the actual chat completion and other factors seem to make no difference one way or the other.

My initial thinking was a misconfigured host server in a server group as the load balancing round robin would produce the effects I was seeing.

Seems clearly a server-side (openAI) issue to me but please LMK if you can ideas or esp. if you use this API and it works well for you.

Thanks
-J

Johns-MacBook-Air:~/src/46a/client/src$ cat foo.sh
curl -vv https://api.openai.com/v1/chat/completions/chatcmpl-BAjzsT1DSPOnwwEERMVrLZEgjBzeL/messages \
  -H "Authorization: Bearer REDACTED" \
  -H "Content-Type: application/json"

Johns-MacBook-Air:~/src/46a/client/src$ sh foo.sh
*   Trying 162.159.140.245:443...
* Connected to api.openai.com (162.159.140.245) port 443
* ALPN: curl offers h2,http/1.1
* (304) (OUT), TLS handshake, Client hello (1):
*  CAfile: /etc/ssl/cert.pem
*  CApath: none
* (304) (IN), TLS handshake, Server hello (2):
* (304) (IN), TLS handshake, Unknown (8):
* (304) (IN), TLS handshake, Certificate (11):
* (304) (IN), TLS handshake, CERT verify (15):
* (304) (IN), TLS handshake, Finished (20):
* (304) (OUT), TLS handshake, Finished (20):
* SSL connection using TLSv1.3 / AEAD-CHACHA20-POLY1305-SHA256
* ALPN: server accepted h2
* Server certificate:
*  subject: CN=api.openai.com
*  start date: Jan 22 23:17:53 2025 GMT
*  expire date: Apr 23 00:17:50 2025 GMT
*  subjectAltName: host "api.openai.com" matched cert's "api.openai.com"
*  issuer: C=US; O=Google Trust Services; CN=WE1
*  SSL certificate verify ok.
* using HTTP/2
* [HTTP/2] [1] OPENED stream for https://api.openai.com/v1/chat/completions/chatcmpl-BAjzsT1DSPOnwwEERMVrLZEgjBzeL/messages
* [HTTP/2] [1] [:method: GET]
* [HTTP/2] [1] [:scheme: https]
* [HTTP/2] [1] [:authority: api.openai.com]
* [HTTP/2] [1] [:path: /v1/chat/completions/chatcmpl-BAjzsT1DSPOnwwEERMVrLZEgjBzeL/messages]
* [HTTP/2] [1] [user-agent: curl/8.4.0]
* [HTTP/2] [1] [accept: */*]
* [HTTP/2] [1] [authorization: Bearer REDEACTED]
* [HTTP/2] [1] [content-type: application/json]
> GET /v1/chat/completions/chatcmpl-BAjzsT1DSPOnwwEERMVrLZEgjBzeL/messages HTTP/2
> Host: api.openai.com
> User-Agent: curl/8.4.0
> Accept: */*
> Authorization: Bearer REDACTED
> Content-Type: application/json
>
< HTTP/2 500
< date: Thu, 13 Mar 2025 23:20:40 GMT
< content-type: application/json
< content-length: 369
< x-request-id: req_646de97b873f489afb4fb7afcccf4551
< openai-processing-ms: 182
< strict-transport-security: max-age=31536000; includeSubDomains; preload
< cf-cache-status: DYNAMIC
< set-cookie: __cf_bm=2cLE6fjkQbNMVuacUJQrKE1CgctNZTEo85aidlSVAEE-1741908040-1.0.1.1-NZZ6957A4jtH1.15U4SGqlRe91JBuxvKJLItvH7e8KxsZY.H7cO1w86WqWUAa34jFd6D3AlbZP1T6sTsXnpSfN9qmfeHeyjftFfZmlfEkzk; path=/; expires=Thu, 13-Mar-25 23:50:40 GMT; domain=.api.openai.com; HttpOnly; Secure; SameSite=None
< x-content-type-options: nosniff
< set-cookie: _cfuvid=69F.lkETL8hTwOJmzKWhwK6tfUCqjGmLO8UPs2vnViQ-1741908040065-0.0.1.1-604800000; path=/; domain=.api.openai.com; HttpOnly; Secure; SameSite=None
< server: cloudflare
< cf-ray: 91ff43603c1c7ad3-SJC
< alt-svc: h3=":443"; ma=86400
<
{
  "error": {
    "message": "The server had an error processing your request. Sorry about that! You can retry your request, or contact us through our help center at help.openai.com if you keep seeing this error. (Please include the request ID req_646de97b873f489afb4fb7afcccf4551 in your email.)",
    "type": "server_error",
    "param": null,
    "code": null
  }
* Connection #0 to host api.openai.com left intact
}Johns-MacBook-Air:~/src/46a/client/src$

Screenshot 2025-03-13 at 4.21.13 PM
Screenshot 2025-03-13 at 4.21.13 PM1920×918 116 KB

2 Likes
2

Is there a round robin on the DNS as well?

I am getting stuff like this for network changes. Maybe that happens somehow? Or are you always requesting from the same IP?

3

Maybe you can try from an isolated server without loadbalancing and see if that still happens and then add one layer after the other?

4

No idea.

My round robin example was pure speculation on my part as the error pattern and context matches exactly the issue profile so looks like that to me. But it appears an openAI API server-side issue AFAIC.

I actually have no idea how open API has their endpoints served.
-J

5

Uff,… I was thinking from the opposite direction as if you were hosting the API… Should get awake first…

Still can you try with an isolated test case? I usually start with a simple curl call from command line.

Err ok… I mean - from another server as your local computer then - to make sure it is not a problem of your machine.

6

Ah wait - shouldn’t that be a post call?

7

No it’s a GET (see screenshot attached to OP).

It works some of the time so that would rules out GET/POST and other stuff.

8

Can you start a docker container and try from inside that?

9

from aws…

[ec2-user@ip-10-0-0-23 ~]$ docker exec -it 18bfc00d1eed /bin/bash
[root@18bfc00d1eed server]# vi foo.sh
[root@18bfc00d1eed server]#
[root@18bfc00d1eed server]# sh foo.sh
* About to connect() to api.openai.com port 443 (#0)
*   Trying 172.66.0.243...
* Connected to api.openai.com (172.66.0.243) port 443 (#0)

and then the same 500

< HTTP/1.1 500 Internal Server Error
< Date: Fri, 14 Mar 2025 01:12:07 GMT
< Content-Type: application/json
< Content-Length: 369
< Connection: keep-alive
< x-request-id: req_1f880801833fad29f98b66c889e4efea
< openai-processing-ms: 350
< strict-transport-security: max-age=31536000; includeSubDomains; preload
< CF-Cache-Status: DYNAMIC
< Set-Cookie: __cf_bm=lr55yepq6Tzzk58VP5Mm76ij_infvvLe6.P6Ix7Ks4c-1741914727-1.0.1.1-BFCYcWtAS0TmD6SjYxs.DuZTqlzmjU6DUBk5UA.D.y_9wkjlttdOR9y0pr89bYpc0xP5G4ur43xly6U41xQwvIznj19J_B5Hlp6_8OWfZQk; path=/; expires=Fri, 14-Mar-25 01:42:07 GMT; domain=.api.openai.com; HttpOnly; Secure; SameSite=None
< X-Content-Type-Options: nosniff
< Set-Cookie: _cfuvid=UtEbqIDjqA5vyf0otM7TH4J1Z6Sd6cpjcdqQhkyOM4Q-1741914727999-0.0.1.1-604800000; path=/; domain=.api.openai.com; HttpOnly; Secure; SameSite=None
< Server: cloudflare
< CF-RAY: 91ffe6a6ca9967f1-SJC
< alt-svc: h3=":443"; ma=86400
<
{
  "error": {
    "message": "The server had an error processing your request. Sorry about that! You can retry your request, or contact us through our help center at help.openai.com if you keep seeing this error. (Please include the request ID req_1f880801833fad29f98b66c889e4efea in your email.)",
    "type": "server_error",
    "param": null,
    "code": null
  }
10

It’s something like:
DNS to Cloudflare geo routing
Cloudflare fingerprinting and DDOS protection
Cloudflare to OpenAI services
Rate limit/auth workers
Inbound to endpoint
API says “sorry about that” with your OpenAI request ID

1 Like