/common directory can be accessed via the
$COMMON environment variable, i.e.
/commonis a network attached FS, so limit the number of files per directory (1 million files in a directory is a very bad idea).
/commonfor a job, you will need to add a line to your submission script!
/commonfor a given job.
To gain access to the path on worker nodes, a job must be submitted with the following SLURM directive:
If a job lacks the above SLURM directive,
/common will not be accessible from the worker nodes. (Briefly, this construct will allow us
to quickly do maintenance on a single cluster without having to unmount
$COMMON from all HCC resources).
/workfor that –
/commonshould mostly be used to read largely static files or data.
/commonis available on machines with different cpu architecture, different network connections, and so on. caveat emptor!
modulethings should be just fine!
The /common file system has the capability to compress files so they store
less data on the underlying disk storage. Tools like
du will report the
true amount of space consumed by files by default. If the files have been
compressed before being stored to disk, the report will appear smaller
than what may be expected. Passing the
--apparent-size argument to
du will cause the report to be the uncompressed size of consumed space.
$ pwd /common/demo/demo01 $ python -c 'print "Hello World!\n" * 2**20,' > hello_world.txt $ ls -lh hello_world.txt -rw-r--r-- 1 demo01 demo 13M Mar 7 12:55 hello_world.txt $ du -sh hello_world.txt 2.0K hello_world.txt $ du -sh --apparent-size hello_world.txt 13M hello_world.txt