2.1.130 under heavy swapping, revisited

Henrik Olsen (henrik@iaeste.dk)
Thu, 3 Dec 1998 08:02:11 +0000 ( )

As a result of reading the tales of doing a make -j modules on a 128MB
ram, 64MB swap machine, I decided to do a small test to see how 2.1.130
reacts in the opposite situation, 64MB ram, 256MB swap, to see if the
problems might have something to do with the reports from earlier
versions of problems with swap<memory.


Pentium 166MHz overclocked to 188MHz, but rock steady for a long time by
now, swap on a Adaptec 1542CP, EIDE IBM DeskStar as main

The test I did was a simple
make clean
make -j modules
while running procinfo and top on different consoles

For reference, I did it under 2.0.36 first, where the result was that the
loadaverage jumped to about 100, freezing top and procinfo for the first
20 minutes, then settled down to about 80 for 1 hour more with about 330
processes running and swap use growing slowly to about 160MB, at which
time finished successfully.

Doing the same under 2.1.130 gave about the same start, except tat the
last message seen by procinfo before it froze had 400+ processes running.

After 6 hours I gave up and stopped the compilation, since procinfo has
segfaulted, top had just stopped, the console had several reports of the
fork: Resource temporarily unavailable
Once stopped, I checked the system for its health, and found that syslog,
named and several other daemons had died during the compile run.

My conclusion is that 2.1.130 still has some problems under really extreme
load, on the other hand, putting the machine under a load that makes it
solve the complete set of problems in about 100 times what it would take
if done serially (make -j 2 modules is over in a few minutes) could be
seen as too silly to bother fixing.

Henrik Olsen,  Dawn Solutions I/S
Get the rest there.

- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu Please read the FAQ at http://www.tux.org/lkml/