-
Slides of the CCPR intro overview of Computing Resources https://ucla.app.box.com/v/slides-stat-comp-pt2
-
CCPR Statistics https://ccpr.ucla.edu/services/statistics-and-methods-core-mission/
-
CCPR Computing https://ccpr.ucla.edu/services/computing/
-
CCN Wiki on Hoffman2 https://www.ccn.ucla.edu/wiki/index.php/Hoffman2
https://www.ccn.ucla.edu/wiki/index.php/Hoffman2:Linux_Tutorial
- NOTE: Multiple commands can fit into a single line by separating commands with a semi colon
- interactive session, 16g ram 1 hour, any available node
qrsh -l i,h_data=16g,h_rt=1:00:00;
- interactive session, ... , using special highpriority CCPR node '1234' (1234 is a fake example)
qrsh -l i,h_data=16g,h_rt=1:00:00,highp -q *@n1234
Using Rstudio
module av R
module load R/3.2.1; module load Rstudio; rstudio
Using Stata (se/mp)
module av stata
module load stata/16; xstata-se
module load stata/15; xstata-mp
For batch jobs, you can use the R.q or stata.q menu helpers to initially generate a template .cmd file.
stata.q
R.q
Use any text editor to further modify the generated .cmd file. Specifically, parallel cores, modify ram/runtime. Look for the line in the .cmd file that looks similar to the below
-l h_data=512M,h_rt=48:00:00,highp -pe shared 8
Using specific node 1234
-l h_data=248G,h_rt=00:5:00 -q *@n1234
find $HOME -name core.'[0-9]*' -print | xargs rm
- NOTE: this removes ALL core files
When a batch job errors out, a huge core dump file is automatically created.
If you are unaware that this core file exists, you might run into hard disk space issues.
Hard disk space issues might prevent GUI apps from loading, say Rstudio.
mygroupjobs
qconf -su ccpr
qstat -u $(qconf -su ccpr)
qstat -q *@n1234
- NOTE: if the node in question is 'n1234'
qstat -u $(qconf -su ccpr) -q *@n1234