Date: Sat, 21 Nov 1998 14:01:03 -0500 From: Dan Root <dar@thekeep.org> To: freebsd-current@FreeBSD.ORG Subject: Memory usage question Message-ID: <19981121140103.A84228@thekeep.org>
next in thread | raw e-mail | index | archive | help
I've got a machine that I recently upgraded from a -current built around
the end of August to the most recent -current. For the most part
everything seems fairly happy: CAM, softupdates, etc.
However I've been having the 'dying daemons' problem. For the most part
I've been working around it, but I just dropped another 128 Megs of RAM
(for a total of 256M) into the machine to help prevent the problem from
occuring.
Needless to say, 2 hours after doing make world, installing a new kernel
and rebooted (under moderately heavy load), I've just touched swap.
With 155M of memory marked as 'inactive' according to top.
It strikes me as a little odd, but why on earth would things be swapping
out if there's 155M of inactive memory? And 8 megs of cache and 8 more
megs of buffers (again, all according to top). This strikes me as overly
agressive swapping.
Along the same lines, why is this memory sitting in the inactive pool
rather than being used for caching or buffering? (Or am I misunderstanding
the meaning of these headings?)
Is there some CONFIG, rc.conf, or sysctl options I have inadvertantly
forgetten to set or tweak?
output from top:
last pid: 13182; load averages: 1.00, 1.13, 1.14 13:45:16
84 processes: 4 running, 79 sleeping, 1 zombie
CPU states: 16.3% user, 0.0% nice, 33.7% system, 1.5% interrupt, 48.6% idle
Mem: 56M Active, 155M Inact, 32M Wired, 8380k Cache, 8344k Buf, 784K Free
Swap: 64M Total, 320K Used, 64M Free
output from vmstat -s:
1471838 cpu context switches
2492270 device interrupts
831953 software interrupts
5341605 traps
11534741 system calls
0 swap pager pageins
0 swap pager pages paged in
5 swap pager pageouts
5 swap pager pages paged out
1840 vnode pager pageins
8805 vnode pager pages paged in
1 vnode pager pageouts
1 vnode pager pages paged out
58 page daemon wakeups
93774 pages examined by the page daemon
977 pages reactivated
1731236 copy-on-write faults
274744 copy-on-write optimized faults
2559046 zero fill pages zeroed
207564 zero fill pages prezeroed
8 intransit blocking page faults
5219033 total VM faults taken
5231071 pages freed
53 pages freed by daemon
124964 pages freed by exiting processes
12642 pages active
40336 pages inactive
2004 pages in VM cache
8032 pages wired down
965 pages free
4096 bytes per page
2748624 total name lookups
cache hits (80% pos + 12% neg) system 0% per-directory
deletions 0%, falsehits 0%, toolong 0%
And of course, the following was syslogged:
swap_pager: suggest more swap space: 511 MB
If anyone would like more information, I'd be happy to provide the output
of sysctl -a, my CONFIG file, hardware listings, or anything else that
might explain what's going on here.
-DaR
--
Dan Root - dar@thekeep.org
To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-current" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?19981121140103.A84228>