Here are the general steps for restarting an attempt to configure the machine.
When the vagrant up doesn't work, before you try again, you need to do the following:
Clear the errored mounts (see below), and stop the machine and destroy it
vagrant suspend
vagrant destroy
Delete any temporary files
sudo rm -rf .vagrant/
I am not sure if this is necessary, but I also deleted the mount folders that were created (eg, image_data, nv_env, and pycortex_data)
For all solutions below, before you try a fresh install (eg vagrant up) you need to delete the errored mount attempts, because they will remain in the system file and continue to trigger errors:
sudo vim /etc/exports
You will see something along the lines of:
# VAGRANT-BEGIN: 1000 9bfb8ce1-3a88-4173-a67e-7b70eff3e94e
"/home/vanessa/Packages/vagrant/neurovault/neurovault_puppet" 192.168.33.10(rw,no_subtree_check,all_squash,anonuid=1000,anongid=wheel,fsid=399783149)
"/home/vanessa/Packages/vagrant/neurovault/nv_env" 192.168.33.10(rw,no_subtree_check,all_squash,anonuid=1000,anongid=wheel,fsid=1116967821)
"/home/vanessa/Packages/vagrant/neurovault/image_data" 192.168.33.10(rw,no_subtree_check,all_squash,anonuid=1000,anongid=wheel,fsid=3802941516)
Delete all the lines that were added by vagrant.
The server should be active on your local machine at 192.168.33.10, but sometimes you will get server errors and an ugly white screen. It is good practice to always restart these servers with any changes, as this is a common fix to the problem:
sudo service uwsgi restart
If you need to debug, logs can be found at `/var/log/nginx' and '/var/log/uwsgi'. I'm not sure if this is useful, but sometimes deleting the neurovault.sock (and having a new one get generated) can help fix issues. I couldn't get it to delete, but moving seemed to work:
mv neurovault.sock /home/vagrant/neurovault.sock.bak
sudo service uwsgi restart
sudo service nginx restart
It's extremely annoying to have the forwarding not work to your local machine, but if you are desperate there is a workaround - to run the django server locally and view on a browser from within the VM (yes, I am wincing as I write this). First, when you ssh with vagrant, you want to make sure to enable X11 forwarding:
vagrant ssh -- -Y
source /opt/nv_env/bin/activate
python /opt/nv_env/NeuroVault/manage.py runserver 8080
This will get the server running, manually. Now you should open a second terminal for firefox:
vagrant ssh -- -Y
sudo apt-get install firefox
firefox
And going to localhost:8080 (or whatever port you have specified) will show the application. You would then need to have a third terminal to do command line stuffs (more wincing).
Make sure that you also have the redis server working:
redis-cli ping
PONG
Count jobs in celery:
redis-cli llen celery
Check celery logs:
cat /var/log/neurovault-tasks/celery.log
You can then try restarting celery and the tasks:
sudo service neurovault-tasks restart
sudo /etc/init.d/neurovault-tasks restart # This may be the same thing, another way
sudo /etc/init.d/celeryev restart
The best way to test your tasks is to start celery manually, which means that you can generate tasks in the web interface and see them complete successfully (or error out) in the window where you are running the server.
./manage.py celery worker -A neurovault.celery -l info -E --autoreload
The similarity metric needs to be added manually, so if you have not done this, you will see this error:
Task neurovault.apps.statmaps.tasks.save_voxelwise_pearson_similarity[b4bc9dd0-a37c-4352-be65-74fe613c78b1] raised unexpected: DoesNotExist('Similarity matching query does not exist.',)
You should add it! You can open up a shell using manage.py, and copy the code from scripts/add_pearson_similarity_metric.py
manage.py shell
>%run scripts/add_pearson_similarity_metric
Sometimes the database is out of data, and you get errors that a relation is missing. You should try making and doing migrations:
/opt/nv_env/NeuroVault/manage.py makemigrations
/opt/nv_env/NeuroVault/manage.py migrate