[Prev][Next][Index][Thread]
SUMMARY - loosing swap
Again, i apologise for late summary. My original post:
>I have a Alpha 3000 300x running OSF/1 v2.0 which is running out off swap.
>It has 64Mb of memory and had 189MB of swap. I have increased this to 250Mb
>in an attempt to stop the problem. I was advised by DEC to increase the swap
>because of this problem and because the screen would "lock" at times for
>periods of thirty seconds to a minute.
>Has anyone found a fix for this or a similar problem?
>
>PS swapon -s shows low use of swap. The probelem seems to come out of the blue
>without any major applications running.
>
Experience from osf1 1.3, may also apply to 2.0 ?
You might like to look at moving from 'greedy' mode to 'lazy' mode. The
default mode has swap space allocated as a process is started, even if
there is no prospect of using it. The 'lazy' mode allocates swap space
only when it is about to be used.
On 1.3 need to get rid of (i.e. rename) /sbin/swapdefault and reboot.
Also reported to help to add -terminate to end of line in the file
/usr/bin/X11/xdm/Xservers. Sounds like some memory leaks in X console
stuff ???
My results pretty good, default system had 60% of swap space used with
5 users logged in, only 13% after changes above.
------------------------------------------------------------
David Bannon D.Bannon@xxxxxxxxxxxxxx
Dept Biochem Phone 61 03 479 2197
Latrobe University Fax 61 03 479 2467
------------------------------------------------------------
-------------------------------------------------------------------------
What is the problem for which you want a fix? The fix for running out of
swap space is to either (a) allocate more swap space, or (b) reduce the
amount needed by the running system. You've already chosen (a).
> PS swapon -s shows low use of swap. The probelem seems to come out of the blue
> without any major applications running.
I've got a DEC 3000 Model 500 with DEC OSF/1 V2.0 running. I have a modest
load of applications fired up most of the time. I'm using about 20% of the
256MB of swap I've got available -- and I'm running with "eager" allocation
(that is, /sbin/swapdefault is in place and pointing to my primary swapping
partition). You MIGHT be able to reduce swap allocation demands by getting
rid of your /sbin/swapdefault symbolic link. However, if you're REALLY out
of space, and REALLY need lots of swap space available, then all that will
happen is something will start failing (unpredictably).
It seems to me that you either are experiencing a memory leak (unlikely, as
the problem corrects itself) or you have something running (in background?)
that you aren't aware of. If it were me, I'd set up a script that I could
run in the background that would run swapon -s and when it find low space,
would get a ps aux into a file for examination. You need to find out what
is eating up all your memory.
Tom
Dr. Thomas P. Blinn
UNIX Software Group
Digital Equipment Corporation
Mailstop ZKO3-3/W20
110 Spit Brook Road
Nashua, New Hampshire 03062
Position: Technical Liaison to Kubota Graphics Corporation
Internet: tpb@xxxxxxxxxxx
Phone: (603) 881-0646
Worry kills more people than work because more people worry than work.
Opinions expressed herein are my own, and do not necessarily represent
those of my employer or anyone else, living or dead, real or imagined.
------------------------------------------------------------------------
There are some things that you can do to reduce the memory usage on an
OSF/1 system. DEC has produced a document called 'Monitoring and Tuning
DEC OSF/1 AXP Systems'. It should be available from your DEC sales
droid.
One of the things we did was to disable save-unders and backing store
from the X11 server. In /usr/lib/X11/xdm/Xservers add '-su -bs
-terminate' to the entry.
Another thing we did was to limit the size of the unified buffer cache by
setting 'ubcmaxpercent 20' in the /sys/conf/{MACHINE} file and rebuilding
the kernel. Note that this change will adversely affect performance in a
machine that does a lot of I/O to local disks.
Before, my 32Mb Alpha was constantly paging, and, after the changes, the
machine is basically useable.
Hope this helps.
Barry Moyer EMail: barrym@xxxxxxxxxxxxxxxxx
Senior Systems Analyst
Valmet Automation
SAGE Systems Division
10333 Southport Road, South West
Calgary Alberta
T2W 3X6 Phone: (403) 253-8848
Canada FAX: (403) 259-2926
----------------------------------------------------------------------------
Hi Dave,
I had a similar problem with two of our new alpha's running V2.0. I even added
secondary swap and ran out of swap at odd times, only to have it released.
It turns out that the problem was caused by the /usr/bin/internet_mom daemon
consuming all of the swap space on the system, grinding it to a halt.
The circumstances that caused this to happen are when a machine's routing table
grows very large (around 200 entries). A netstat -rn will show you the
routing table....
In any case, a call to Colorado Support solved the problem. They provided me
with new copies of /BINARY/kinfo.o and /BINARY.rt/kinfo.o - a kernel
rebuild was then required.
Hope this helps with your problem.
Kathleen Berry
Phillips Laboratory, Hanscom AFB
berry@xxxxxxxxxxxxx
__________________________________________________________________________
> PS swapon -s shows low use of swap. The probelem seems to come out of the blue
> without any major applications running.
Have you set swap to over commitment mode or are you still swapping
save? Alphas tend to use enormous amounts of swap, if swapping save
(the default). To switch to over commitment mode, rename or remove
the file /etc/swapdefault and reboot. This fixed all swapspace
problems for us here.
Kristian
UNIX should be a household word.
---------------------------------------------------------------------------
Dear David:
Two things you should know if you don't already:
1. There is "lazy" swap and "eager" swap. The latter reserves all the
swap space a process might ever need, thereby using it up very
quickly. You can see this by the difference in Reserved and In-use
space using swapon -s. To eliminate this, rm the link
/sbin/swapdefault
2. xdm continues to reserve more memory each time somebody logs in.
To stop this edit the line :0 local /usr/bin/X11/X adding -terminate
to the end in /usr/lib/X11/xdm/Xservers
Peter Stern
Chemical Physics Department
Weizmann Institute of Science
76100 Rehovot, ISRAEL
----------------------------------------------------------------------------
Hi Dave!
I am having a similar problem. However I have not yet proved to myself
that its a swap problem. Two of my Alpha 3000/800 have a full 1GB ram
and 1.5GB swap space. We run large programs and I don't like to spend
time paging/swaping. The same thing happens, everything locks up for
(in my case upto 5 min.) then frees up and all is well again!
If you get any clues as to why this is, please drop me a line!
Thanks
Marc Cozzi
Univ. of Notre Dame cozzi@xxxxxx
----------------------------------------------------------------------------
See the release notes for the 2.0 OSF, look for the note on swapspace;
you have to add a -terminate option to the window manager.
If you dont have'em or cant find'em, send me anote back and i'll look for mine.
Steve Schneider
-------------------------------------------------------------------------
Hmmmmm, seems to be a commmon problem. I have tried some of these solutions
without success but I'm not finished yet. The -terminate for xdm is well
documented, and I did this from day one. What I can't fathom is that I set
up two new machines, a 3600 and a 3300, both the same (except for disk space
and physical memory), the 600 is fine and the 300 is crap!
When I find what fixes this I'll let y'all know.
THANKS
--
David Robson (ROBBO),BHP Information Technology, P.O.Box 21,
Whyalla South Australia 5600, Ph: +61 86 404596 Fax:...404720,
E-mail dave@xxxxxxxxxxxxxxxxx
"Life's battle's don't always go
To the stronger or faster man,
But soon or late, the man who wins,
IS THE MAN WHO THINKS HE CAN."
-- Walter W. Wintle