Disk space policy: Difference between revisions

From DISI
Jump to navigation Jump to search
No edit summary
(asdf)
 
(5 intermediate revisions by 2 users not shown)
Line 1: Line 1:
This is the lab disk space policy and applies to those with ssh access to our cluster only.  This policy currently applies to [[cluster 2]]. It will apply to [[cluster 0]] with caveats by June 1, 2014.
This is the lab disk space policy and applies to those with ssh access to our cluster only.  This policy currently applies to [[cluster 2]].
We ask you to observe the following disk usage policy for your account on the cluster. We think this policy will cover 90% of the users, 90% of the time.  We do understand that research can have unanticipated needs, and we are willing to work with you to create a plan that works for everyone.
 
Please observe the following disk usage policy for your account on the cluster. This policy should cover 90% of users, 90% of the time.  We understand that research can have unanticipated needs, and we are willing to work with you to create a plan that works for everyone.


{{TOCright}}
{{TOCright}}
= Summary =
We back up cluster home directories daily, and selected database, crystallography and other archived files quarterly. We do not back up /nfs/work files and we do not back up workstations or laptops. To backup files from /nfs/work, repatriate the key files to your home directory whence they will be backed up as part of normal procedures. To backup laptop and desktop, use USB disks, which we can give you.
Our backup systems cannot keep up with the growth in disk space and the voracious appetite for disk of group members. This is a pragmatic policy that, while requiring a little organization, is more likely to deliver restored files you can actually rely on.
As always, we welcome [[feedback]].
{| class="wikitable"
|-
! Name !!  Location !! Quota !! Backup?  !! Usage
|-
| Home || /nfs/home/<userid> || 500 GB || Near-daily || General use - not large database and not docking runs.
|-
| Work || /nfs/work/<userid> || 500GB* || Never || Docking runs or calculations that produce large output
|-
| DB  || /nfs/db/<userid> || 2TB*  || Quarterly at best  || Database files, often persistent, rarely modifed
|-
| Store || /nfs/store/<userid> || 500GB* || Quarterly at best || Crystallography files or other archived files, rarely modified.
|-
| Desktop || /home/<userid> || N/A || Never || Copy critical files from desktop to home directory on server.
|-
| Laptop || / || N/A || Never || Copy critical files from laptop to home directory on server.
|}
* * More space may be requested from the [[sysadmin]]s.


= Checking Disk Usage =
= Checking Disk Usage =
You can check your disk usage on the cluster NFS (excluding /nfs/db) drives by running the `quota` command on any cluster machine.
You can check your disk usage on the cluster NFS (excluding /nfs/db) drives by running the '''quota''' command on any cluster machine.


Example Output:
Example Output:
Line 17: Line 44:
                       114M    500G    512G              3      0      0         
                       114M    500G    512G              3      0      0         
     nfs_store:/export/store/
     nfs_store:/export/store/
                           4    500G    512G              1      0      0    
                           4    500G    512G              1      0      0


= Treatment by disk type =
= Treatment by disk type =
When you log in to a cluster computer, you are in your home directory.  This would normally be /nfs/home/<userid>. Lab tradition is to have a ~/code/ directory for all software and a ~/work/ directory for all of your research. Your home directory is backed up weekly, and has a quota of 500 GB. Keeping the quota at this level will allow us to provide a high performance backup system that runs regularly and completes in a timely fashion.
When you log in to a cluster computer, you are in your home directory.  This would normally be /nfs/home/<userid>. Lab tradition is to have a ~/code/ directory for all software and a ~/work/ directory for all of your research. Your home directory is backed up weekly, and has a quota of 500 GB. Keeping the quota at this level will allow us to provide a high performance backup system that runs regularly and completes in a timely fashion. /nfs/home is backed up around 300 days a year. Use it!


== Docking jobs ==
== Docking jobs ==
If you require more than your quota for ~/work/, we will create a special directory for you, e.g. /nfs/work/<userid>, which we suggest you symlink to ~/work/.  We expect most lab members engaged in multiple docking projects will require space on /nfs/work.  /nfs/work will never be backed up. We recommend you repatriate key files from time to time from /nfs/work/<userid> to e.g. ~/worksave/. We can help you write scripts to automate this. You are also welcome to make your own supplementary backups to a USB drive on your workstation or laptop. We have USB drives you may have for this purpose.
If you require more than your quota for ~/work/, we will create a special directory for you, e.g. /nfs/work/<userid>, which we suggest you symlink to ~/work/.  We expect most lab members engaged in multiple docking projects will require space on /nfs/work.  /nfs/work will NEVER be backed up. We recommend you repatriate critical files from time to time from /nfs/work/<userid> to e.g. ~/worksave/. We can help you write scripts to automate this. We encourage you to make your own supplementary backups to a USB drive on your workstation or laptop. We have USB drives you may have for this purpose.


Historical note: in the past, we used to attempt to backup nearly every file on every disk, including millions of docking job output files that had a transient existence. This resulted in slow computers - due to the backup process - and a lack of clarity of what was backed up.  It was unsatisfactory.  One problem has been clarity: how to understand what is backed up and what is not backed up.  By backing up only /nfs/home/<userid> we are forcing you to choose the key input files from which you can re-create your output.  You may also wish to use /nfs/store/<userid> to retain important results. But we want to get away from the idea of backing up every file on every disk, particularly those of transient existence and marginal value.
Fun historical note: In the past, we used to attempt to backup nearly every file on every disk, including millions of docking job output files that had a transient existence. This resulted in slow computers - due to the backup process - and a lack of clarity of what was backed up.  It was unsatisfactory.  One problem has been clarity: how to understand what is backed up and what is not backed up.  By backing up only /nfs/home/<userid> we are forcing you to choose the key input files from which you can re-create your output.  You may also wish to use /nfs/store/<userid> to retain important results. But we want to get away from the idea of backing up every file on every disk, particularly those of transient existence and marginal value.


== Dockable database files ==
== Dockable database files ==
If you create dockable database files that persist and are larger than your quota, ask us to create dedicated space for you, e.g. /nfs/db/<userid>.  The hallmarks of database files is that they are large, they are often used by more than one person, and they are persistent.  Database files are backed up quarterly beginning on the first day of the quarter.
If you create dockable database files that persist and are larger than your quota, ask us to create dedicated space for you, e.g. /nfs/db/users/<userid>.  The hallmarks of database files is that they are large, they are often used by more than one person, and they are persistent.  Database files are backed up quarterly beginning on the first day of the quarter.
 
 
[[Category:Policy]]
[[Category:Internal]]


== Large datasets, completed projects ==
== Large datasets, completed projects ==
If you have large files from the synchrotron or elsewhere, or a project has finished but you want to keep it online, ask us to create dedicated space for you, e.g. /nfs/store/<userid>.  Storage files are backed up quarterly beginning on the first day of the quarter.  
If you have large files from the synchrotron or elsewhere, or a project has finished but you want to keep it online, ask us to create dedicated space for you, e.g. /nfs/store/<userid>.  Storage files are backed up quarterly beginning on the first day of the quarter.  


If after using /nfs/db, /nfs/work and /nfs/store to offload from /nfs/home/ you still need more space, please work with the [[sysadmin]]s to help you.
If after using /nfs/db, /nfs/work and /nfs/store to offload from /nfs/home/ you still need more space, please work with the [[sysadmin]]s.


= Workstation and Laptop =
= Workstation and Laptop =
When you log in to your workstation, you may be in a local directory on the workstation. Our policy is to never back up workstations. You may back up your own workstation, and you may make additional backups of your data on the cluster by writing to USB disks mounted on your desktop or laptop.  We offer as a parting gift from the lab two multi-TB drives to which you may copy your files.  Make one copy to take with you and one copy for us to keep safe for you. Label them clearly in ink. Then, delete the files from the server as you leave the lab, provided the project has been published.
When you log in to your workstation, you may be in a local directory on the workstation. Our policy is to never back up workstations. You may back up your own workstation, and you may make additional backups of your data on the cluster by writing to USB disks mounted on your desktop or laptop.  We offer as a parting gift from the lab two multi-TB drives to which you may copy your files.  Make one copy to take with you and one copy for us to keep safe for you. Label them clearly in ink. Then, delete the files from the server as you leave the lab, provided the project has been published.
= Summary =
We back up home directories weekly, and selected database, crystallography and other archived files quarterly. We do not back up /nfs/work files and we do not back up workstations or laptops. To backup files from /nfs/work, you must repatriate the key files to your home directory whence they will be backed up as part of normal procedures. To backup laptop and desktop, use USB disks, which we can give you.
Our backup systems cannot keep up with the growth in disk space and the voracious appetite for disk of group members. This is a pragmatic policy that, while requiring a little organization, is more likely to deliver restored files you can actually rely on.
As always, we welcome [[feedback]].
{| class="wikitable"
|-
! Name !!  Location !! Quota !! Backup?  !! Usage
|-
| Home || /nfs/home/<userid> || 500 GB || Weekly || General use - not large database and not docking runs.
|-
| Work || /nfs/work/<userid> || 512GB* || Never || Docking runs or calculations that produce large output
|-
| DB  || /nfs/db/<userid> || 2TB*  || Quarterly  || Database files, often persistent, rarely modifed
|-
| Store || /nfs/store/<userid> || 512GB* || Quarterly || Crystallography files or other archived files, rarely modified.
|-
| Desktop || /home/<userid> || N/A || Never || Copy critical files from desktop to home directory on server.
|-
| Laptop || / || N/A || Never || Copy critical files from laptop to home directory on server.
|}
* More space may be requested from the [[sysadmin]]s.


= See Also =  
= See Also =  
* [[Backups]]
* [[Backups]]


 
[[Category:Policy]]
 
[[Category:Internal]]
[[Category:Internal]]

Latest revision as of 04:06, 3 June 2015

This is the lab disk space policy and applies to those with ssh access to our cluster only. This policy currently applies to cluster 2.

Please observe the following disk usage policy for your account on the cluster. This policy should cover 90% of users, 90% of the time. We understand that research can have unanticipated needs, and we are willing to work with you to create a plan that works for everyone.

Summary

We back up cluster home directories daily, and selected database, crystallography and other archived files quarterly. We do not back up /nfs/work files and we do not back up workstations or laptops. To backup files from /nfs/work, repatriate the key files to your home directory whence they will be backed up as part of normal procedures. To backup laptop and desktop, use USB disks, which we can give you.

Our backup systems cannot keep up with the growth in disk space and the voracious appetite for disk of group members. This is a pragmatic policy that, while requiring a little organization, is more likely to deliver restored files you can actually rely on.

As always, we welcome feedback.

Name Location Quota Backup? Usage
Home /nfs/home/<userid> 500 GB Near-daily General use - not large database and not docking runs.
Work /nfs/work/<userid> 500GB* Never Docking runs or calculations that produce large output
DB /nfs/db/<userid> 2TB* Quarterly at best Database files, often persistent, rarely modifed
Store /nfs/store/<userid> 500GB* Quarterly at best Crystallography files or other archived files, rarely modified.
Desktop /home/<userid> N/A Never Copy critical files from desktop to home directory on server.
Laptop / N/A Never Copy critical files from laptop to home directory on server.
  • * More space may be requested from the sysadmins.


Checking Disk Usage

You can check your disk usage on the cluster NFS (excluding /nfs/db) drives by running the quota command on any cluster machine.

Example Output:

   [teague@gimel ~]$ quota -s
   Disk quotas for user teague (uid 42001): 
        Filesystem  blocks   quota   limit   grace   files   quota   limit   grace
   nfs_home:/export/home/
                      321M    500G    512G             472       0       0        
   nfs_work:/export/work/
                      114M    500G    512G               3       0       0        
   nfs_store:/export/store/
                         4    500G    512G               1       0       0

Treatment by disk type

When you log in to a cluster computer, you are in your home directory. This would normally be /nfs/home/<userid>. Lab tradition is to have a ~/code/ directory for all software and a ~/work/ directory for all of your research. Your home directory is backed up weekly, and has a quota of 500 GB. Keeping the quota at this level will allow us to provide a high performance backup system that runs regularly and completes in a timely fashion. /nfs/home is backed up around 300 days a year. Use it!

Docking jobs

If you require more than your quota for ~/work/, we will create a special directory for you, e.g. /nfs/work/<userid>, which we suggest you symlink to ~/work/. We expect most lab members engaged in multiple docking projects will require space on /nfs/work. /nfs/work will NEVER be backed up. We recommend you repatriate critical files from time to time from /nfs/work/<userid> to e.g. ~/worksave/. We can help you write scripts to automate this. We encourage you to make your own supplementary backups to a USB drive on your workstation or laptop. We have USB drives you may have for this purpose.

Fun historical note: In the past, we used to attempt to backup nearly every file on every disk, including millions of docking job output files that had a transient existence. This resulted in slow computers - due to the backup process - and a lack of clarity of what was backed up. It was unsatisfactory. One problem has been clarity: how to understand what is backed up and what is not backed up. By backing up only /nfs/home/<userid> we are forcing you to choose the key input files from which you can re-create your output. You may also wish to use /nfs/store/<userid> to retain important results. But we want to get away from the idea of backing up every file on every disk, particularly those of transient existence and marginal value.

Dockable database files

If you create dockable database files that persist and are larger than your quota, ask us to create dedicated space for you, e.g. /nfs/db/users/<userid>.  The hallmarks of database files is that they are large, they are often used by more than one person, and they are persistent. Database files are backed up quarterly beginning on the first day of the quarter.

Large datasets, completed projects

If you have large files from the synchrotron or elsewhere, or a project has finished but you want to keep it online, ask us to create dedicated space for you, e.g. /nfs/store/<userid>. Storage files are backed up quarterly beginning on the first day of the quarter.

If after using /nfs/db, /nfs/work and /nfs/store to offload from /nfs/home/ you still need more space, please work with the sysadmins.

Workstation and Laptop

When you log in to your workstation, you may be in a local directory on the workstation. Our policy is to never back up workstations. You may back up your own workstation, and you may make additional backups of your data on the cluster by writing to USB disks mounted on your desktop or laptop. We offer as a parting gift from the lab two multi-TB drives to which you may copy your files. Make one copy to take with you and one copy for us to keep safe for you. Label them clearly in ink. Then, delete the files from the server as you leave the lab, provided the project has been published.

See Also