-
MemFree is misleading
Hello everybody ,
i’ve started the memfree alarm for a debian agent . I noticed that alarm is triggered although i see only a few process on server .
I think that command in agent is not correctly : “cat /proc/meminfo | grep MemFree | awk ‘{ print $2 }'” because memory management in linux is not critical if this value is low .I suggest “free -m | grep buffers/cache | awk ‘{print $4}'”
What do you think ?
-
5 Replies
-
-
Hi again.
I test in my machine with run a virtualbox (this program use much memory) and without this program:
md@speccy:~$ cat /proc/meminfo | grep MemFree MemFree:Â Â Â Â Â 33120 kB md@speccy:~$ cat /proc/meminfo | grep MemFree MemFree:Â Â Â Â 1274172 kB md@speccy:~$
And the value go down when the machine have less memory.
-
-
Ok. I have talk with jobmates and you are right. Because the memory in GNU/Linux is management in other priorities and the system may take all memory but it is free.
I have add this changes in the “Pending Task” tracker of project in the SourceForge:
https://sourceforge.net/tracker/?func=detail&aid=3284422&group_id=155200&atid=1226163
But at the moment you could change the line for you correct line in your system.
Thanks.
-
Hi guys.
 I use this modules to monitor the memory, and they are pretty usefull to me.
———————————————————————————————–
# Memory Session
module_begin
module_name memfree
module_type generic_data
module_exec free |grep “buffers/cache” | awk {‘print $4’}
module_description Free memory
module_endmodule_begin
module_name memused
module_type generic_data
module_exec free |grep “buffers/cache” | awk {‘print $3’}
module_description Used memory
module_endmodule_begin
module_name mem_comm
module_type generic_data
module_exec cat /proc/meminfo | grep “Committed_AS:” | awk ‘{ print $2 }’
module_description Commited memory
module_endmodule_begin
module_name mem_swpin
module_type generic_data
module_exec cat /home/pandora/swp.txt |grep “Average” | awk ‘{ print $2 }’
module_description Total swap pages in per second
module_endmodule_begin
module_name mem_swpout
module_type generic_data
module_exec cat /home/pandora/swp.txt |grep “Average” | awk ‘{ print $3 }’
module_description Total swap pages out per second
module_end
———————————————————————————————–My alerts are set based on sar -W (for page faults). This is used when the memory is over and the system starts to swaping (using virtual memory).
The script I use to run sar and generate the swp.txt is cpu.sh:
———————————————————————————————–
#!/bin/bash
# Versao 1.0
# Script para recuperar a media de processamento
# dos ultimos 5 minutos.
MIN=$(date +%M)
HOR=$(date +%H)
MSA=$(date -d “5 minutes ago” +%M)
HSA=$(date -d “5 minutes ago” +%H)
CPU=$(echo “/home/pandora/cpu.txt”)
SWP=$(echo “/home/pandora/swp.txt”)# O script repete o ultimo valor obtido a meia noite
if [ $HOR -eq 0 ] && [ $MIN -lt 5 ]; then
    #Nao atualiza os arquivos cpu e swp e finaliza script
    cat $CPU
    exit
fi# Ajuste para uso do sar com valores com 2 digitos
INICIO=$(printf “%s:%s:00” $HSA $MSA)
FIM=$(printf “%s:%s:00” $HOR $MIN)sar -s $INICIO -e $FIM > $CPU
sar -W -s $INICIO -e $FIM > $SWP
cat $CPU
———————————————————————————————–I call the script at the first module in pandora, so the result in cpu.txt and swp.txt are ready for the other modules.
———————————————————————————————–
# SAR Session
module_begin
module_name cpu_sar_user
module_type generic_data
module_exec /home/pandora/cpu.sh | grep Average |awk ‘{ print $3 }’
module_description CPU sar used by Users
module_end
———————————————————————————————–This script has a flaw for the first value of the day, but this is not a problem to me.
I hope this helps. Sorry for the loooong post 😉
Regards.
