[bug] Failed to fetch repository fatal: not a git repository (or any parent up to mount point /)

[gaeljw]

Describe the bug

Our Sourcebot instance has a couple of repositories that appear in error in the UI.

After looking at the logs, I can see the following error message:

[backend] | 2025-08-06T15:04:50.961Z �[32minfo�[39m: [repo-manager] Fetching group/project...
[backend] | 2025-08-06T15:04:50.968Z �[31merror�[39m: [repo-manager] Failed to sync repository gitlab.mycompany.net/group/project (id: 345) after 3 attempts. Error: Error: Failed to fetch repository /data/repos/345: fatal: not a git repository (or any parent up to mount point /)
[backend] | Stopping at filesystem boundary (GIT_DISCOVERY_ACROSS_FILESYSTEM not set).
[backend] | 
[backend] | 2025-08-06T15:04:50.973Z �[32minfo�[39m: [repo-manager] Repo index job for repo group/project (id: 345, jobId: 1049966) failed with error: Error: Failed to fetch repository /data/repos/345: fatal: not a git repository (or any parent up to mount point /)
[backend] | 
[backend] | Stopping at filesystem boundary (GIT_DISCOVERY_ACROSS_FILESYSTEM not set).

If I look in the data/repos/345, it's effectively missing files and appear to not be a Git repository. But why?

/data/repos/345 # ls -al
total 28
drwxr-xr-x    2 nobody   nobody        4096 Jul  1 08:15 .
drwxr-xr-x  777 nobody   nobody       20480 Jul 31 07:43 ..
-rw-r--r--    1 nobody   nobody           0 Jul  1 08:15 FETCH_HEAD

The GitLab repository exists and is accessible from the token used.

Apparently the files haven't been modified since July 1st. I wonder if something wrong happened at that time and since then it's not able to "recover" from it?

Maybe in case of such error, Sourcebot could force to clear and start over for this repo?

I tried the "Retry index" button in the UI but it didn't help.

To reproduce

I don't know :|

Sourcebot deployment information

Sourcebot version (e.g. v3.0.1): 4.6.1

Additional information

No response

[gaeljw]

For the record, deleting the folder /data/repos/345 seems to have helped.

[brendan-kellam]

Hmm maybe the repository was marked for garbage collection, and then during garbage collection it got interrupted and got into this weird state? I'm not for sure though - it will be tough to debug 🥹

[gaeljw]

This sounds like a possible explanation indeed.

No worries, I don't expect you to troubleshoot this one :) It was more as a record for anyone encountering it as well. Looks like cleaning the repository cache is good enough for when this happens.

[Brumstar]

I have this issue as well with the following configuration -

{
    "$schema": "https://raw.githubusercontent.com/sourcebot-dev/sourcebot/main/schemas/v3/index.json",
    "connections": {
        "gitlab-connection": {
            "type": "gitlab",
            "url": "https://gitlab.custom.net/",
            "all": true,
            "token": {
                "env": "GITLAB_TOKEN"
            }
        }
    }
}

From the logs, every single repository failed to sync, with the error

[backend] | 2025-09-12T12:47:10.302Z error: [repo-manager] Failed to sync repository gitlab.custom.net/repo (id: 69), attempt 2/3. Sleeping for 120s... Error: Error: Failed to fetch repository: /data/.sourcebot/repos/69. Reason: fatal: not a git repository (or any parent up to mount point /)

Note sure what I've done wrong here. This is going through first time setup.

[brendan-kellam]

@Brumstar does this consistently happen every time? Could you try clearing your sourcebot cache and trying again?

[Brumstar]

Hi @brendan-kellam what is the process for clearing the cache? I will try it but it seemed to happen again on another fresh deploy.

[brendan-kellam]

@Brumstar You can delete the .sourcebot folder in the directory that you mounted to the docker container.

It's strange that the fetch is failing because it implies that the clone succeeded.

If you don't mind, could you try the following steps:

1. Delete the .sourcebot folder to clear the cache

2. Modify your config such that a) only a single project will be indexed, and b) set the connection reindex interval to 1 minute (so you don't need to wait an hour, the default)

{
    "$schema": "https://raw.githubusercontent.com/sourcebot-dev/sourcebot/main/schemas/v3/index.json",
    "settings": {
        "reindexIntervalMs": 60000 // 1 minute
    },
    "connections": {
        "gitlab-connection": {
            "type": "gitlab",
            "url": "https://gitlab.custom.net/",
            "projects": [
                "my-group/my-project"
            ],
            "token": {
                "env": "GITLAB_TOKEN"
            }
        }
    }
}

3. Start the instance. I'm wondering if we will hit this error for this single project or not.

[Brumstar]

Interestingly after deleting the cache and retrying, things worked in both single project and all project configurations. I'm not sure how possible it will be to debug my replication of this issue.

[brendan-kellam]

Hmm ok, glad it is working now.. still not sure what the root cause is 😅 Will keep this issue open in case others run into it.