I am on a shared Unix server and I want to run several Stata jobs in batch mode. I have a loop that generates these batch jobs and executes them. The problem is that all the Stata batch processes are executed simultaneously, overloading the server and depriving my colleagues of much needed memory.
I want to find out how much memory the Unix system has at a given time, then execute if there's enough memory available. I wrote this piece of code to handle that, using Stata-MP 13.1 on Unix
is there another command to get the available memory in a Unix system within Stata?
If it doesn't exist, I will submit this simple program to SSC. Could anyone help with a solution for my freemem program that uses file read instead of using preserve and import delimited to read the vmstat output?
Thank you.
I want to find out how much memory the Unix system has at a given time, then execute if there's enough memory available. I wrote this piece of code to handle that, using Stata-MP 13.1 on Unix
Code:
clear
* Program to determine amount of free memory
cap program drop freemem
program define freemem, rclass
preserve
quietly {
cap erase free.txt
- Vmstat gives available memory
shell (vmstat)>free.txt
clear
import delimited free.txt, delimiters("whitespace") rowrange(3:3) varnames(nonames)
keep v1
split v1
keep v14
li
loc fmem = v14[1]
loc fmem = `fmem'/1e6
noi di "`fmem'"
return local fmem=`fmem'
restore
}
end
* Test; Run regression only if memory exceeds a value
loc exceed=164.7
foreach i in a b c d{
loc run=0
while `run'==0 {
freemem
loc fm=r(fmem)
if `fm'>`exceed' {
di "Available mem = `fm' - clear to run"
sysuse auto, clear
reg price mpg
loc run=1
}
else {
di "Memory not available, waiting ..."
pause 1000
}
}
}
If it doesn't exist, I will submit this simple program to SSC. Could anyone help with a solution for my freemem program that uses file read instead of using preserve and import delimited to read the vmstat output?
Thank you.
Comment