[SLUG] The mystery of the missing memory!
If I run top and hit M to get a listing sorted by memory usage, I see that 98% of memory is used up, as follows: 20% buffers 20% cached 23% non 0.0% processes 1% free Where is the missing 36%? Ah! I think to myself I have 52 processes other processes using '0.0%' memory. What if I guess they're each really using 0.04% memory Rats, that only explains another 2.6%. Any ideas? I'm stumped. $ top --version top (procps version 2.0.7) This is on an old RH 7.2 system using kernel 2.4.28. top interesting bits: 11:13pm up 74 days, 56 min, 1 user, load average: 0.10, 0.04, 0.01 84 processes: 77 sleeping, 2 running, 2 zombie, 3 stopped CPU states: 0.0% user, 0.3% system, 0.0% nice, 99.6% idle Mem: 255716K av, 251356K used,4360K free, 0K shrd, 53528K buff Swap: 265032K av, 237716K used, 27316K free 52428K cached PID USER PRI NI SIZE RSS SHARE STAT %CPU %MEM TIME COMMAND 1273 root 18 0 255M 10M 732 R 0.0 4.3 1012m X 2451 luke 18 0 13932 8864 2388 S 0.0 3.4 7:21 postilion.exec 1061 xfs9 0 16780 7492 336 S 0.0 2.9 0:49 xfs 17278 stella 9 0 6440 6440 4092 S 0.0 2.5 0:00 postilion.exec 15573 stella 9 0 6432 6432 4088 S 0.0 2.5 0:17 postilion.exec 15571 root 9 0 2236 2184 1480 S 0.0 0.8 0:07 sshd 5937 luke 9 0 12192 2096 1160 S 0.0 0.8 0:00 postilion.exec 745 ntp9 0 1924 1924 1732 S 0.0 0.7 0:00 ntpd 16392 root 9 0 1644 1644 1256 S 0.0 0.6 0:00 fetchmail 17160 stella 9 0 1416 1412 400 S 0.0 0.5 0:00 ispell 1290 luke 9 0 1868 1344 684 S 0.0 0.5 2:06 wmaker 18057 luke 9 0 2948 1164 428 S 0.0 0.4 105:16 smbd 17321 luke 18 0 968 968 756 R 0.1 0.3 0:00 top 1309 luke 9 0 1108 944 644 S 0.0 0.3 0:02 bash 5918 luke 9 0 1356 808 368 S 0.0 0.3 0:00 ispell 1307 luke 9 0 936 724 428 S 0.0 0.2 0:01 bash 1295 luke 9 0 1948 648 296 S 0.0 0.2 0:07 wterm 1296 luke 9 0 1956 640 352 S 0.0 0.2 0:09 wterm 625 rpcuser9 0 664 568 564 S 0.0 0.2 0:00 rpc.statd 1084 root 9 0 880 564 428 S 0.0 0.2 0:04 nmbd 572 root 9 0 608 552 512 S 0.0 0.2 0:03 syslogd 15901 root 9 0 508 508 436 S 0.0 0.1 0:00 portsentry 577 root 9 0 1352 496 452 S 0.0 0.1 0:00 klogd 15897 root 9 0 492 492 420 S 0.0 0.1 0:01 portsentry 597 rpc9 0 572 488 484 S 0.0 0.1 0:00 portmap 800 root 9 0 452 440 252 S 0.0 0.1 0:13 smartd 1297 luke 9 0 584 428 328 S 0.0 0.1 0:00 xcb 947 root 8 0 848 424 308 S 0.0 0.1 0:06 sendmail 10423 luke 9 0 668 420 300 S 0.0 0.1 0:00 wterm 818 root 9 0 528 400 336 S 0.0 0.1 0:04 sshd 1299 luke 9 0 468 364 260 S 0.0 0.1 45:07 wmmail 1079 root 9 0 688 296 208 S 0.0 0.1 0:00 smbd 874 lp 8 0 360 252 208 S 0.0 0.0 0:00 lpd 10424 luke 8 0 592 240 236 S 0.0 0.0 0:00 bash 1298 luke 9 0 296 228 176 S 0.0 0.0 0:10 wmclock 852 root 9 0 328 140 136 S 0.0 0.0 0:00 xinetd 10616 luke 9 0 220 13292 S 0.0 0.0 0:03 asmixer 985 root 9 0 172 11676 S 0.0 0.0 0:00 crond 903 root 9 0 200 112 108 S 0.0 0.0 0:00 rpc.rquotad 533 root 9 0 168 108 104 S 0.0 0.0 0:00 mingetty 1142 root 8 0 136 8060 S 0.0 0.0 0:00 rhnsd 1 root 8 0 120 7252 S 0.0 0.0 0:06 init 1121 daemon 9 0 104 4432 S 0.0 0.0 0:00 atd 1087 root 9 0 456 3632 S 0.0 0.0 0:00 nmbd 1764 root 9 0 156 2824 S 0.0 0.0 0:00 rpc.mountd 1201 root 9 0 268 2016 S 0.0 0.0 0:00 login 1203 root 9 076 1612 S 0.0 0.0 0:00 mingetty 1204 root 9 076 1612 S 0.0 0.0 0:00 mingetty 1205 root 9 076 1612 S 0.0 0.0 0:00 mingetty [... the remaining 35 0.0%-memory using process omitted] I also notice that if I start 5 galeon browser windows, I get 6 galeon processes each using 34.4% memory (another 200% memory used). The man page for top says %MEM The task's share of the physical memory. None of these numbers add up, for me! Even if I assume the man page is wrong, and it should say %MEM The task's share of the physical memory, except when the process that's listed is sharing memory with others. Puzzled, luke -- SLUG - Sydney Linux User's Group Mailing List - http://slug.org.au/ Subscription info and FAQs:
Re: [SLUG] The mystery of the missing memory!
There's a nice thread on the gentoo forum that explains this (and some more) http://forums.gentoo.org/viewtopic.php?t=175419 On 2/10/06, Luke Kendall [EMAIL PROTECTED] wrote: If I run top and hit M to get a listing sorted by memory usage, I see that 98% of memory is used up, as follows: 20% buffers 20% cached 23% non 0.0% processes 1% free Where is the missing 36%? Ah! I think to myself I have 52 processes other processes using '0.0%' memory. What if I guess they're each really using 0.04% memory Rats, that only explains another 2.6%. Any ideas? I'm stumped. $ top --version top (procps version 2.0.7) This is on an old RH 7.2 system using kernel 2.4.28. top interesting bits: 11:13pm up 74 days, 56 min, 1 user, load average: 0.10, 0.04, 0.01 84 processes: 77 sleeping, 2 running, 2 zombie, 3 stopped CPU states: 0.0% user, 0.3% system, 0.0% nice, 99.6% idle Mem: 255716K av, 251356K used,4360K free, 0K shrd, 53528K buff Swap: 265032K av, 237716K used, 27316K free 52428K cached PID USER PRI NI SIZE RSS SHARE STAT %CPU %MEM TIME COMMAND 1273 root 18 0 255M 10M 732 R 0.0 4.3 1012m X 2451 luke 18 0 13932 8864 2388 S 0.0 3.4 7:21 postilion.exec 1061 xfs9 0 16780 7492 336 S 0.0 2.9 0:49 xfs 17278 stella 9 0 6440 6440 4092 S 0.0 2.5 0:00 postilion.exec 15573 stella 9 0 6432 6432 4088 S 0.0 2.5 0:17 postilion.exec 15571 root 9 0 2236 2184 1480 S 0.0 0.8 0:07 sshd 5937 luke 9 0 12192 2096 1160 S 0.0 0.8 0:00 postilion.exec 745 ntp9 0 1924 1924 1732 S 0.0 0.7 0:00 ntpd 16392 root 9 0 1644 1644 1256 S 0.0 0.6 0:00 fetchmail 17160 stella 9 0 1416 1412 400 S 0.0 0.5 0:00 ispell 1290 luke 9 0 1868 1344 684 S 0.0 0.5 2:06 wmaker 18057 luke 9 0 2948 1164 428 S 0.0 0.4 105:16 smbd 17321 luke 18 0 968 968 756 R 0.1 0.3 0:00 top 1309 luke 9 0 1108 944 644 S 0.0 0.3 0:02 bash 5918 luke 9 0 1356 808 368 S 0.0 0.3 0:00 ispell 1307 luke 9 0 936 724 428 S 0.0 0.2 0:01 bash 1295 luke 9 0 1948 648 296 S 0.0 0.2 0:07 wterm 1296 luke 9 0 1956 640 352 S 0.0 0.2 0:09 wterm 625 rpcuser9 0 664 568 564 S 0.0 0.2 0:00 rpc.statd 1084 root 9 0 880 564 428 S 0.0 0.2 0:04 nmbd 572 root 9 0 608 552 512 S 0.0 0.2 0:03 syslogd 15901 root 9 0 508 508 436 S 0.0 0.1 0:00 portsentry 577 root 9 0 1352 496 452 S 0.0 0.1 0:00 klogd 15897 root 9 0 492 492 420 S 0.0 0.1 0:01 portsentry 597 rpc9 0 572 488 484 S 0.0 0.1 0:00 portmap 800 root 9 0 452 440 252 S 0.0 0.1 0:13 smartd 1297 luke 9 0 584 428 328 S 0.0 0.1 0:00 xcb 947 root 8 0 848 424 308 S 0.0 0.1 0:06 sendmail 10423 luke 9 0 668 420 300 S 0.0 0.1 0:00 wterm 818 root 9 0 528 400 336 S 0.0 0.1 0:04 sshd 1299 luke 9 0 468 364 260 S 0.0 0.1 45:07 wmmail 1079 root 9 0 688 296 208 S 0.0 0.1 0:00 smbd 874 lp 8 0 360 252 208 S 0.0 0.0 0:00 lpd 10424 luke 8 0 592 240 236 S 0.0 0.0 0:00 bash 1298 luke 9 0 296 228 176 S 0.0 0.0 0:10 wmclock 852 root 9 0 328 140 136 S 0.0 0.0 0:00 xinetd 10616 luke 9 0 220 13292 S 0.0 0.0 0:03 asmixer 985 root 9 0 172 11676 S 0.0 0.0 0:00 crond 903 root 9 0 200 112 108 S 0.0 0.0 0:00 rpc.rquotad 533 root 9 0 168 108 104 S 0.0 0.0 0:00 mingetty 1142 root 8 0 136 8060 S 0.0 0.0 0:00 rhnsd 1 root 8 0 120 7252 S 0.0 0.0 0:06 init 1121 daemon 9 0 104 4432 S 0.0 0.0 0:00 atd 1087 root 9 0 456 3632 S 0.0 0.0 0:00 nmbd 1764 root 9 0 156 2824 S 0.0 0.0 0:00 rpc.mountd 1201 root 9 0 268 2016 S 0.0 0.0 0:00 login 1203 root 9 076 1612 S 0.0 0.0 0:00 mingetty 1204 root 9 076 1612 S 0.0 0.0 0:00 mingetty 1205 root 9 076 1612 S 0.0 0.0 0:00 mingetty [... the remaining 35 0.0%-memory using process omitted] I also notice that if I start 5 galeon browser windows, I get 6 galeon processes each using 34.4% memory (another 200% memory used). The man page for top says %MEM The task's share of the physical memory. None of these numbers add up, for me! Even if I assume the man page is wrong, and it should say
Re: [SLUG] The mystery of the missing memory!
On Monday 13 February 2006 02:39, [EMAIL PROTECTED] wrote: I also notice that if I start 5 galeon browser windows, I get 6 galeon processes each using 34.4% memory (another 200% memory used). The man page for top says %MEM The task's share of the physical memory. None of these numbers add up, for me! Even if I assume the man page is wrong, and it should say %MEM The task's share of the physical memory, except when the process that's listed is sharing memory with others. You need to forget your windows-centric view of memory as a commodity. Try xosview to see the global view of your memory usage. Processes ALWAYS share memory with others (but eg: ash is built to not use shared libs). Some processes show memory usage that is actually swap. Imagine the confusion: John runs top on TheProg and finds its memory usage goes from 100M down to 10M as swap gets used. Jane runs top on TheProg and finds its memory usage goes from 100M to 250M !! Top lets me examine a process, not (try) to do system accounting Frankly there is no such thing as 90% memory used. 90% of memory should ALWAYS be used, if not you have a faulty system. So I guess the question should be 'what are you trying to measure'. For me your galeon example is exactly what I'd expect to see. This is useful info, and no doubt exactly what top is designed to display. James -- SLUG - Sydney Linux User's Group Mailing List - http://slug.org.au/ Subscription info and FAQs: http://slug.org.au/faq/mailinglists.html
