Hello,
Since yesterday morning I have been receiving GPT-4 API errors practically every time I send a query. Sometimes it happens that the first query will go through, but subsequent queries keep receiving errors like the one here:
Error: Request timed out: HTTPSConnectionPool(host='api.openai.com', port=443): Read timed out. (read timeout=600). Retrying in 5 seconds...
Error: Request timed out: HTTPSConnectionPool(host='api.openai.com', port=443): Read timed out. (read timeout=600). Retrying in 5 seconds..
Right now i’m getting timouts or overloaded errors for every request.
Do you have same probblems? It’s API problem ? Is there any option to solve it?
Quite a few of us are getting problems, it’s quite unreliable in the last few days. If you’re not getting timed out you may be getting 429s. There are a few other topics on it in this section. We cannot do much but wait.
My hypothesis is that it’s because of the new plugins rollout, many people are using GPT-4.
Thanks :). I’m most getting timout, somteimtes 429.
Same intermittent “read timed out” errors here.
I’d like to know how to catch these errors and retry etc.
To catch the errors, you use the exception handling features of your language.
If it’s python; wrap the call in a try: except: clause, for example.
Note that you should look at the error for specifically the errors you know how to re-try. There are many other kinds of errors that can happen where re-trying won’t help, and will just make your integration flakier or burn API credits.
Effectively re-trying on errors is a pretty deep subject, that a single forum post probably isn’t the right spot to teach…
thanks @jwatte
Yes I do have a decorator with try/catch + retry + expo-backoff in my python code.
The key as you said is which exceptions to catch for the retries.
I was hoping I can catch a generic openai.error.OpenAIError, but even with that, some failures leak through.
I have same problem. Even you catch error, that’s wont help. You probably will have “timeout error” or 502 erorr, which is caused by OPEN AI API overload.
Error: Request timed out: HTTPSConnectionPool(host='api.openai.com', port=443): Read timed out. (read timeout=600). Retrying in 5 seconds...
You can try many different ways to solve it… I have same script which work perfect in the past, after they publish plugins, api is timing out all the time with the same script.
I’m using GPT-4 and it llooks like it’s overloaded 
from 5-7 days.
1 Like
I have the same problem with the API (GPT4)
Same problem, gpt4 stayed timeout error since midnight
Using gpt-3.5-turbo here… constant timeouts, it’s unusable at the moment. 10+ retries in a row with a 60-second timeout. Having to tell customers that it’s OpenAI, even though the status page doesn’t say there’s an issue.
Same here, my app was working fine up until a couple days ago, now getting:
Retrying langchain.chat_models.openai.ChatOpenAI.completion_with_retry.<locals>._completion_with_retry in 60.0 seconds as it raised Timeout: Request timed out: HTTPSConnectionPool(host='api.openai.com', port=443): Read timed out.