"fatal: not a git repository"

[rosano]

I have been pulling an external git repository via n8n into the container's /tmp folder and have pushed changes successfully via an active workflow. At some point the workflow fails because of a fatal: not a git repository error and I'm not sure why this happens.

When navigating to the .git folder via the container's web terminal in Cloudron, the data inside looks like this:

total 104
drwxr-xr-x  8 cloudron cloudron  4096 May 14 12:00 .
drwxr-xr-x  7 cloudron cloudron  4096 May 14 12:00 ..
-rw-r--r--  1 cloudron cloudron     5 May 14 07:08 COMMIT_EDITMSG
-rw-r--r--  1 cloudron cloudron    92 May 14 07:08 FETCH_HEAD
-rw-r--r--  1 cloudron cloudron    41 May 14 07:08 ORIG_HEAD
drwxr-xr-x  2 cloudron cloudron  4096 May  3 10:17 branches
-rw-r--r--  1 cloudron cloudron   333 May 14 07:08 config
drwxr-xr-x  2 cloudron cloudron  4096 May 14 12:00 hooks
-rw-r--r--  1 cloudron cloudron 56770 May 14 07:08 index
drwxr-xr-x  2 cloudron cloudron  4096 May 14 12:00 info
drwxr-xr-x  3 cloudron cloudron  4096 May  3 10:18 logs
drwxr-xr-x 52 cloudron cloudron  4096 May 14 07:08 objects
drwxr-xr-x  5 cloudron cloudron  4096 May  3 10:18 refs

When cloning/pulling the repo fresh to a new folder, it looks slightly different but I'm not sure what would make it behave differently.

total 116
drwxr-xr-x  8 cloudron cloudron  4096 May 19 07:10 .
drwxr-xr-x  7 cloudron cloudron  4096 May 19 07:10 ..
-rw-r--r--  1 cloudron cloudron     5 May 19 07:10 COMMIT_EDITMSG
-rw-r--r--  1 cloudron cloudron    92 May 19 07:10 FETCH_HEAD
-rw-r--r--  1 cloudron cloudron    23 May 19 07:10 HEAD
-rw-r--r--  1 cloudron cloudron    41 May 19 07:10 ORIG_HEAD
drwxr-xr-x  2 cloudron cloudron  4096 May 19 07:10 branches
-rw-r--r--  1 cloudron cloudron   333 May 19 07:10 config
-rw-r--r--  1 cloudron cloudron    73 May 19 07:10 description
drwxr-xr-x  2 cloudron cloudron  4096 May 19 07:10 hooks
-rw-r--r--  1 cloudron cloudron 57218 May 19 07:10 index
drwxr-xr-x  2 cloudron cloudron  4096 May 19 07:10 info
drwxr-xr-x  3 cloudron cloudron  4096 May 19 07:10 logs
drwxr-xr-x 14 cloudron cloudron  4096 May 19 07:10 objects
-rw-r--r--  1 cloudron cloudron   114 May 19 07:10 packed-refs
drwxr-xr-x  5 cloudron cloudron  4096 May 19 07:10 refs

I found one answer suggesting to copy .git/ORIG_HEAD to .git/HEAD in the broken repo, but that didn't change anything and creates another fatal: detected dubious ownership in repository at XXXXX error.

As it functions for weeks and then suddenly stops, I wonder if it's not an n8n issue and something that happens at the Cloudron or OS level. If there are any guesses about that I would appreciate it.

For now I just manually delete the folder and pull again, but this defeats the purpose of using an automation system; I can also delete as part of the workflow, but wanted to avoid pulling a large amount of data everytime.

[nebulon]

Relying on /tmp folder as persistent data is generally not a good idea. Tmp storage cleanup script might run at any time as well as if app containers are recreated the tmpfs volume is not guaranteed to stay around.

Have you tried to use a subfolder in /app/data/ for this? This is where data is kept persistently in Cloudron.

[girish]

@rosano said in "fatal: not a git repository":

At some point the workflow fails because of a fatal: not a git repository error and I'm not sure why this happens.

The HEAD file in git can never go missing. It could be that a previous workflow or command just crashed midway for some reason. I think you have to check the git/workflow logs to figure what's going on. But the core issue is that HEAD file should never disappear in a git repo .

[rosano]

@nebulon said in "fatal: not a git repository":

Have you tried to use a subfolder in /app/data/ for this? This is where data is kept persistently in Cloudron.

Been using this folder and seems like it has had no issues for a week or two now thanks for clarifying