Using Temporary Storage on CCR Servers
Temporary "scratch" space is available to users in several locations on the clusters for use during job execution. Things you should know about scratch space on CCR servers:
This temporary space is not backed up and is subject to being automatically removed after it reaches one month old.
- For fastest I/O, users can make use of local disk space in /scratch on all compute nodes.
- This data is automatically deleted at the end of the running job to make room for the next job that runs on that node.
- Do NOT use /tmp or /temp on any compute nodes as it will fill up the root drive and crash the node, killing your job and any others running on the node!
- For additional space, users also have access to /gpfs/scratch
- Automatic removal of files older than 30 days takes place on a daily basis in /gpfs/scratch so please make sure to only use this as temporary storage.
- You will also be asked to remove files if you are taking up more than your fair share of space in this filesystem.
- IMPORTANT NOTICE: Your user account privileges will be revoked if you make attempts to get around this rule (i.e. scripting the touching of files to keep them current). If you have any questions about the use of these scratch spaces, please contact us via CCR help.
More information on storage, backups and data retention: