...
| Excerpt |
|---|
These policies ensure that every user can efficiently take advantage of Pawsey filesystems. |
| Column |
|---|
|
| Warning |
|---|
- IMPORTANT: Purge policy on
/scratch will change since Monday 10th June 2024. The window of inactivity of files before purge will be reduced to 21 days (instead of the previous rule of 1-month). Files will be automatically deleted after exceeding the allowed window of inactivity. Please read carefully the policies on this page and stick to the rules and best practices for Pawsey users.
|
|
Staff access to project directories
...
When raising a request for help, users can choose not to let helpdesk staff access project directories by clearly stating in the ticket that access to project files is not permitted. Note that this could increase the time taken to resolve technical queries or limit the supercomputing experts' ability to assist.
/scratch
...
usage
The /scratch filesystem should be used for working data, which is input and output files actively used by jobs queued or running on the supercomputer.
To prevent performance impacts on other users, there are limits of 1 PB capacity per project and 1 million files per user on /scratch.
You can check your usage of the /scratch filesystem in terms of number of files and kilobytes. Run the following command:
...
| Column |
|---|
|
| Code Block |
|---|
| language | bash |
|---|
| theme | DJango |
|---|
| title | Terminal 1. Check usage of the /scratch quota |
|---|
| $ lfs quota /scratch
Disk quotas for user aelwell (uid 20701):
Filesystem kbytes quota limit grace files quota limit grace
/scratch 1460 0 0 - 14 0 0 - |
|
Best practices
| Column |
|---|
|
| Tip |
|---|
- Data should be moved out of the
/scratch filesystem regularly, into longer-term storage such as Acacia or third-party storage facilities. For this, check /wiki/spaces/DATA/pages/54459526.
|
| Tip |
|---|
- The
/scratch filesystem is a shared resource that suffers low performance when the number of files increases too much. Users should actively remove their files if no longer in use, rather than waiting for the system to delete them.
|
|
...
| Tip |
|---|
- By default,
/scratch use PFL to stripe files to make best use of the SSD and HDD. This default settings has been set to work for almost all workflows and we recommend not setting the stripping explicitly. For this, check Progressive File Layout (PFL) and File Striping.
|
| Warning |
|---|
- Users should never use the
touch command to avoid purging of their files. The use of the touch command to avoid purging generates overloading of the metadata servers of the /scratch filesystem. Remember that /scratch is shared among all users. Therefore, users should respect the best practices to avoid overloading of metadata servers at all times. Overloading of the metadata servers dramatically degrade performance affecting all users at the same time. Pawsey reserves the right to revoke access to users that do not respect this best practice.
|
|
Purge policy of 21 days
Files on the /scratch filesystem are deleted automatically after 21 days from their last access (reading or modifying). This action is referred to as purging. It is part of a dynamic allocation system Pawsey implements to maximise the availability of space for large-scale, data-intensive computations.
In exceptional circumstances, if the /scratch filesystem is close to being full, files that are less than 30 21 days old may be deleted. However, in such circumstances, Pawsey contacts users by email to advise of the situation and to give them time to transfer their data files to other storage facilities.
Data should be moved out of the scratch filesystem as soon as is practically possible, into longer-term storage such as Acacia or third-party storage facilities.
| Column |
|---|
| Info |
|---|
With a purge policy, users are able to use more of the scratch system when they need it. For example, if projects A and B each require 500 TB of space to run one big job, Project A can use that space at one time, and project B can use that same disk space at another time (when the data for Project A has been removed by the user or purged by the system). Under an allocation policy, these two projects together would need to be allocated half of the available resources, leaving little for the rest of the projects on the machine. Many other supercomputing facilities also utilise a scratch purge policy, including CSCS (Swiss National Supercomputing Centre), NERSC (Lawrence Berkeley Laboratory, USA) and OLCF (Oak Ridge Leadership Computing Facility, USA). |
|
...
Related pages
...
Other supercomputing facilities with purge policies