Corrupted git on launcher rebuild app

Product Support Self-hosting

AI-generated summary

Anisus encountered a corrupted Git repository error during ./launcher rebuild app after upgrading docker_manager via the UI, which caused the Admin dashboard to render blank. The error originated inside the app container at /var/www/discourse, showing repeated “packfile does not match index” messages. Attempts to fix it by manually cloning a fresh Discourse repo inside the container worked temporarily, but the issue resurfaced during rebuild. Merefield confirmed the issue was container-internal and noted there are two Git layers (one outside the container). Ultimately, anisus resolved the issue by stopping the app, pruning all Docker containers and images, and rebuilding — which also triggered a PostgreSQL upgrade. A final rebuild completed successfully, restoring full functionality. The root cause appeared to be a combination of Git corruption and lingering artifacts from a prior Docker upgrade.

1

After upgrading docker_manager through the UI (which seemed to go fine), the Admin dashboard just rendered an empty page, even if the rest of the forum seemed to work.

To resolve this, I resorted to run:

./launcher rebuild app

However, during that process, it ran into errors:

I, [2025-07-14T08:54:06.062789 #1]  INFO -- : > cd /var/www/discourse && sudo -H -E -u discourse bash -c '
  set -o errexit
  if [ $(git rev-parse --is-shallow-repository) == "true" ]; then
      git remote set-branches --add origin main
      git remote set-branches origin tests-passed
      git fetch --depth 1 origin tests-passed
  else
      git fetch --tags --prune-tags --prune --force origin
  fi
'
error: packfile .git/objects/pack/pack-581fe47ace192793da04d01989a4d1a7f0e6103c.pack does not match index
error: packfile .git/objects/pack/pack-581fe47ace192793da04d01989a4d1a7f0e6103c.pack does not match index
error: packfile .git/objects/pack/pack-581fe47ace192793da04d01989a4d1a7f0e6103c.pack does not match index
error: packfile .git/objects/pack/pack-581fe47ace192793da04d01989a4d1a7f0e6103c.pack does not match index

… and the error: packfile just continues to be spammed until interrupted.

From the looks of it, the git repository has become corrupted. But what git repository?
I don’t even have a /var/www/discourse folder, so I assume this is inside a container volume?

I am a bit stumped. Thanks for helping!

2

What do you mean by that?

Normally you’d update docker_manager via the UI.

3

Aah! My mistake. I meant UI, not API. I’ll edit it right way.

4

I tried to have resolved it by starting and entering the app container:

./launcher start app
./launcher enter app

Then I verified that the error occurred when calling the git command as the user discourse:

cd /var/www/discourse
sudo -H -E -u discourse bash -c 'git fetch --tags --prune-tags --prune --force origin'

Resulting in:

error: packfile .git/objects/pack/pack-581fe47ace192793da04d01989a4d1a7f0e6103c.pack does not match index
error: packfile .git/objects/pack/pack-581fe47ace192793da04d01989a4d1a7f0e6103c.pack does not match index
error: packfile .git/objects/pack/pack-581fe47ace192793da04d01989a4d1a7f0e6103c.pack does not match index
etc..

I tried to resolve it by renaming discourse and clone a new one:

cd /var/www
mv discourse discourse-old
git clone https://github.com/discourse/discourse.git
chown -R discourse:discourse discourse

Once cloned, tried to run the git fetch again with no errors!

cd /var/www/discourse
sudo -H -E -u discourse bash -c 'git fetch --tags --prune-tags --prune --force origin'

But as I exited the container shell and tried to rebuild again with ./launcher rebuild app, I got the same error.

Does the ./launcher rebuild app use a different container than ./launcher enter app enters?

5

yes, there are two levels of git, one outside (not including all the cloned plugins!)

your error does indeed appear to occur inside the container … most strange!

6

Finally got it to work!
I think it was a mixture between a corrupted git and the result of a previous docker upgrade. I resolved it by pruning all docker containers and images:

cd /opt/discourse
./launcher stop app
docker container prune
docker image prune -a
./launcher rebuild app

This resulted/ended with:

Upgrade Complete
----------------
Optimizer statistics are not transferred by pg_upgrade.
Once you start the new server, consider running:
    /usr/lib/postgresql/15/bin/vacuumdb --all --analyze-in-stages

Running this script will delete the old cluster's data files:
    ./delete_old_cluster.sh
-------------------------------------------------------------------------------------
UPGRADE OF POSTGRES COMPLETE

Old 13 database is stored at /shared/postgres_data_old

To complete the upgrade, rebuild again using:

./launcher rebuild app
-------------------------------------------------------------------------------------

By running the rebuild again, it started working fully:

./launcher rebuild app