I have a codex environment that I had previously set up and used successfully, including the interactive terminal in the edit page for the environment. The last time I tried the interactive terminal was 1-2 months ago. Today I tried opening the interactive terminal, and it consistently fails to start, with a non-descript “Terminal errored - An unexpected error occurred” message.
Checking the network traffic, when starting the environment, I see a 200 POST to https://chatgpt.com/backend-api/wham/environments/test which streams back the following response:
I am suspecting some backend issue at OpenAI, though I’m not seeing any relevant indications on OpenAI’s status page currently.
I do not immediately have another env to test, in case it is some peculiar issue with this particular env. Though again, this worked previously and to my knowledge has been unchanged since then.
Post-setup processing is also failing to establish any net connection - was previously working to authorized sites, no related config changes, this is on the Codex side.
As with several of the above posters, this is resolved for me now as well, without any changes on my end. So indeed must have been some temporary issue on OpenAI’s end.
Unclear if some users might still be experiencing issues, as the above poster reported the same issue after others reported it had been fixed.
I have also been experiencing this. So have other members of my orgs Teams account. It has been intermittent yesterday, but today has been for the entire working day.
I’m also running into this. The interactive terminal was working for me on Monday, but has been completely non-functional yesterday and today with this error.
I’ve been having the same issue. I’ve been unable to use Codex for three days now. Any environment I create, if I try to start a test container it just flashes “starting test” for a split second before throwing me a “Terminal Error - An unexpected error occurred”
To be more precise, issues started around noon (UTC+2) Tuesday on July 22, with sputtering service. Codex tasks would get aborted early, because at some point the container used by Codex would crash out. Given the average time a task takes this caused it to happen essentially always (usually within 5 minutes)
Outage got worse with containers periodically refusing to start outright later in the day. Throughout the whole day on July 23, and right now on July 24th, 9 AM UTC+2, it hasn’t been able to fire the container up anymore, and unfortunately only gives a generic server error message directly after trying to start:
When streaming the above response the crash with the second message comes give or take 5 seconds after the first one.
Typical steps like removing and readding the Github connect in Codex and in Github, recreating repositories, even blank ones just to see if launching the test container will work, have yielded zilch yet.
If it makes a difference, I just tried to run a test container in environment settings, and the process stayed up long enough to get a heartbeat in. It still failed before initializing the container.
I would guess that tasks in essence work for me as well, if I would have a test container that can survive the task long enough. The issues I got on the first day didn’t kill the container before even starting it up - it cloned the repo and worked through its task as usual before eventually the container would crash out and consequently the task would fail.
No updates on this thread for 20 days. I still have the same problem: my Terminal fails every time after “Test complete”. Has anyone found a workaround to share?
