Getting generic "Failed to set up container" error

Coding with ChatGPT
1

Just getting this generic error after a bit.

Starting your container
Downloading repo
Failed to set up container.

no retry or specifics of what is failing exactly.

2

I had this happen the other day on a repo with a lot of directories (>1000).

What is your repo situation? Also, have you defined and committed your AGENTS.md for the repo yet?

3

I had this happen today on an environment that was working for several tasks, not sure what happened, if it was transient or due to something in the repo once it downloaded. I rolled back the last change and it started working. You may know this, but if you open the environment for your repo here https://chatgpt.com/codex/settings/environments and go to advanced, there is a test terminal and if you click it it will trigger an environment build live for you with full terminal output that may be more insightful than what you get in the task digest.

Screenshot 2025-05-19 at 8.41.52 PM
Screenshot 2025-05-19 at 8.41.52 PM2716Ă—1202 470 KB

4

This also happens to me, and also on a very large repo (many thousands of files). It would, at minimum, be nice if we were allowed the option for a partial checkout.

5

FYI this was happening for me and then when I decided to test locally and did a fresh clone of my repo… it failed to clone even on my machine (some issues with LFS).

So, try setting up on local first and see what errors you get there.

Also if you’re getting failures in the later steps here’s a handy script for setting up a local docker image that’s as similar as possible to the OpenAI VM.

1 Like
6

Thanks for pointing this out, LFS was the issue on my end as well.

After setting
GIT_LFS_SKIP_SMUDGE=1
in the Environment variables section of the codex environment allowed the clone to go through in the terminal inside the codex environment editor, but the clone still fails when launching a task.

Here’s how I got it to work:

  1. Create a <dummy-repo> in GitHub
  2. Obtain a GitHub personal access token
  3. Create a codex environment for <dummy-repo>
  4. Set Env Var GIT_LFS_SKIP_SMUDGE to 1
  5. Set Secret GH_PAT to the PAT you acquired
  6. Use the following setup script:
git lfs install --skip-smudge

-x # Avoid echoing PAT (probably better way to handle this)
ENC=$(printf 'x-access-token:%s' "$GH_PAT" | base64)
git -c http.extraHeader="Authorization: Basic $ENC" \
    clone \
        --branch <branch> \
        --single-branch \
        --depth 1 \
        --filter=blob:none \
        --sparse \
        https://github.com/<org>/<big-repo>.git \
        /workspace/<dummy-repo>/<big-repo>
 
cd /workspace/<dummy-repo>/<big-repo>
git -c http.extraHeader="Authorization: Basic $ENC" sparse-checkout init --cone
git -c http.extraHeader="Authorization: Basic $ENC" sparse-checkout set <dir to checkout>
+x

Note that this will not work for “Ask” queries. Only “Code” queries. I think because “Ask” queries do not run the setup scripts.

2 Likes
7

Happy it worked out for you! I ended up expunging LFS from my repo :stuck_out_tongue:

1 Like