Bug #1908

System looses the memory with time

Added by zhtw almost 6 years ago. Updated almost 6 years ago.

Target version:
Start date:
Due date:
% Done:



It seams that with time the total amount of memory I see in top is getting less
and less with time until the system hangs.

The system hanged a couple of times and I didn't know why, but recently I
noticed that the system uses swap while I have 6G of memory and don't really do
anything with it. Top showed the total amount of memory less than I really have.
I rebooted and got to see the normal amount.

uname -a:
DragonFly 2.8-RELEASE DragonFly v2.8.2.16.gfc7bf0-RELEASE #7:
Fri Nov 5 16:15:04 MSK 2010
:/usr/obj/usr/src/sys/X86_64_GENERIC x86_64

dmesg.boot is in attach

I don't know what information is needed to trace the bug. Please ask. Thank you.

dmesg.boot (10.3 KB) zhtw, 11/12/2010 09:51 AM


#1 Updated by zhtw almost 6 years ago

On Sun, 14 Nov 2010 16:46:54 +0000, "Matthew Dillon \(via DragonFly
issue tracker\)" <> wrote:

> Matthew Dillon <> added the comment: The first
> thing to do is to compile up the latest 2.8.x kernel in the release branch.
> Your freezes are probably related to a console deadlock that we have fixed.
> On swap use and memory use, we don't have enough information. 90% of the bug
> reports related to memory use turn out to be misinterpretations of system
> statistics, since DragonFly (like all BSDs) use all available free memory to
> cache file data and drop the cache as memory is needed. If swap is growing
> unbounded over time and you are seeing constant paging activity then it could
> be a runaway process running on your machine, which you should be able to
> locate using miscellanious system tools such as 'ps axl'.
> -Matt

Thank you for your answer.

I am alreay using v2.8.2.16.gfc7bf0-RELEASE. Isn't it the latest 2.8.x

Maybe you are right about misinterpretation of memory
statistics, because what I see in "top" now is:

== top ==
Memory: 2978M Active, 1482M Inact, 1171M Wired, 215M Cache, 605M Buf,
62M Free
== top ==

Total amount is 6513M which is more than I have:

== dmesg.boot ==
real memory = 6440680448 (6142 MB)
avail memory = 6037917696 (5758 MB)
== dmesg.boot ==

So now I'm not sure whether the number less than 6G indicates that
is wrong ether.

Anyway I started to collect memory and swap statistics with munin,
so that next time it hangs I will see what happened in details.

Aleksej Lebedev

#2 Updated by zhtw almost 6 years ago

Memory is not the problem.

The system still hangs within ~2 days of use. I'll open another issue.

Also available in: Atom PDF