Skip to main content

Download files

The Files tab is the right place to pull data across jobs. To download the output of a single job, work from its Results page instead. See download your results.

From the Files tab​

  1. Open the Files tab and select the space holding what you want.
  2. Narrow the list with the filters described below.
  3. Select the files or folders, then click Download.

Selecting more than one file compresses them first, which can add a few minutes before the transfer starts.

Narrow the list first, then select and Download. The filters available depend on which space you are in.

Narrow the list before you select. Date and extension filters are in the folder view; filtering by type or owner needs the legacy view, which is what you want when you're after every output file across a month of jobs.

Single-file and folder limits​

These caps decide whether the browser is viable at all:

LimitValue
Single download attempt1 GB
View a file in the browserUnder 1 MB
API single-file download5 GB

Above 1 GB, use the Transfer Manager or the CLI. A browser download has no resume. An interruption at 90% means starting over.

Folder downloads flatten the directory structure

Downloading a folder gives you a flat list of files, not the tree.

This matters for multi-run jobs, where several runs can produce files with the same name. They will collide. Download runs individually, or use the Transfer Manager, which preserves structure.

From a job that's still running​

You don't have to wait for a job to finish.

  • Snapshot captures every file in a run at a point in time, from the Active Runs table. See live tailing and snapshots.

  • scp copies files off the cluster directly, if you've configured SSH access. Job files live in ~/work/:

    scp -i ~/.ssh/<private-key> <cluster-id>:~/work/process_output.log ~/local/path/

    Reverse the arguments to push a file to the cluster. Wildcards work: velocity-contour-*.

    Pushing a file this way puts it on the cluster only. Rescale Files never sees it, so it won't appear in your job folder or your library, and it won't survive the cluster.

    Both the username and the host come from the job's own logs. Look for the line that begins Instance ready for ssh access and copy them out of the command it prints:

    Instance ready for ssh access. Command: ssh -i </path/to/key> uprod.iFJTvf@44.204.104.152

    On Windows, WinSCP does the same over SFTP. Set the file protocol to SFTP, the host name and user name to the values from that line, and leave the password empty, since you authenticate with the key.

Bulk and scripted downloads​

Past a single job's results, the right tool depends on whether you are matching a pattern, looping over jobs, or waiting for runs to finish:

You wantUse
Everything from one job, fastTransfer Manager
Specific files matched by patternTransfer Manager filters
Many jobs, scriptedRescale CLI, looping over job IDs
Output as soon as jobs finishTransfer Manager automatic downloads

Consider not downloading at all​

If you're going to post-process the data anyway, attach the job to a workstation and work with the files in place. No transfer, and you get visualization hardware. For a large result set this is usually faster.

Troubleshooting​

Most download failures are one of four things: