This page lists common problems with storage, grouped by what you were doing, each with its cause and fix.
They were on the container disk, which is erased when an instance stops or restarts and when a Run Job ends. Only network volumes keep files. Mount one at /workspace for work you need to keep. See Choose where to keep each file.
Deleting a volume erases its data, and it can't be recovered. Keep a copy of data you can't recreate, for example by downloading it with featherless-cloud volume cp -r vol://my-data/results ./results.
HF_HOME=/workspace/hf for Hugging Face models, and install packages with pip install --no-cache-dir. To make the container disk larger, edit the instance; the restart erases the container disk, so copy anything you need first.df -h /workspace and increase its capacity. See The volume is full.Check its usage on the Storage page or with df -h /workspace inside an instance, then increase its capacity. Running instances see the new size right away. See Resize a volume.
Volumes are set when you deploy and can't be added or removed afterwards, even with the instance stopped. Copy anything you need off the old instance's container disk, then deploy a new instance with the volume mounted. To deploy with the same settings, save the old instance's settings as a template first.
Several instances and Run Jobs can write to one volume at once, and nothing stops two of them writing the same file. Give each its own output folder, for example by including the job's name in the path.
Instances still mount it. Stop them first, then use Detach and delete, which removes the volume from them before deleting it.
A new volume takes about a minute to be ready for files. Wait, then upload again.
The console uploads files up to 64 GiB. Use the CLI for larger files:
featherless-cloud volume cp ./model.safetensors vol://my-data/models/
Run the same volume cp or volume sync command again. It resumes where it stopped; add --update to skip files that already match.
If you're still stuck, ask us on Discord. Include: