Mount a bucket on an instance or Slurm Cluster

Mount a bucket on an instance, a Virtual Machine, or a Slurm Cluster to read its data as a folder. A bucket always mounts read-only, and many workloads can mount the same bucket at once.

Before you begin

• A bucket in the Ready state, since only a Ready bucket can be mounted. See Create a bucket.
• An instance, a Virtual Machine, or a Slurm Cluster in the same project and region as the bucket. A Slurm Cluster also needs the bucket’s access key; see Find a bucket’s access keys and endpoint.
• Your work saved on the instance. Mounting on a running instance restarts it.
• A place for your results. The mount is read-only, so save results with an S3 client and the bucket’s keys. You can also save them on a Block Storage disk, a Shared File System, or a Parallel File System mounted beside it.

Mount on an instance or Virtual Machine

You can pick buckets when you create the instance or Virtual Machine: open Storage > Other storage > Datasets on the create form. To mount a bucket on an instance or Virtual Machine that already exists:

1. Open the instance’s page and select the Associated Datasets tab. It lists the project’s datasets with their Storage Type, EOS Bucket or Disk.
2. Tick the bucket, select Update, and confirm. A running instance restarts.

On an instance or Virtual Machine, the bucket appears at /datasets/<alias>, where <alias> is the bucket’s Alias (Dataset Name). For the full steps and the mount path rules, see Mount datasets and file systems on an instance.

Mount on a Slurm Cluster

A Slurm Cluster mounts a bucket at a path you choose. Set the path next to the dataset in Dataset Storage on the create form, or on the cluster’s Volumes tab, under Datasets. A Slurm path starts with /, contains no // or spaces, and does not end with /. On the Volumes tab, choose a path other than /home/jovyan, /mnt/local, /datasets, or anything under /datasets/. A Slurm Cluster can mount only bucket datasets. See Mount and unmount volumes.

Check the mount

On an instance, the Associated Datasets tab lists the bucket, and /datasets/<alias> holds its objects. A write to that folder reports that it is read-only, which is expected.

You can empty or delete a mounted bucket only after you unmount it. When you delete an instance, the buckets mounted on it remain. If the bucket is missing from the workload, see Troubleshooting TIR Object Storage.