On a server, with less than 10 users, my server load stays at 1.00 all
the time. It seems that the qmlnxd is very heavy on the cpu. How can I
optimize that and avoid this server load ?

There is a AMD Athlon dual core 1900, with 1Go.

top - 15:24:29 up 109 days, 5 min,  8 users,  load average: 1.00,
1.04, 1.05
Tasks: 163 total,   2 running, 161 sleeping,   0 stopped,   0 zombie
Cpu(s): 24.6% us, 25.6% sy,  0.0% ni, 48.5% id,  1.3% wa,  0.0% hi,
0.0% si
Mem:    895108k total,   863336k used,    31772k free,    73984k
buffers
Swap:   503800k total,    12052k used,   491748k free,   628000k
cached

I am not sure what is going on there, but we have 71 users online and
here is my load:

top - 10:41:00 up 33 days, 22:32, 71 users,  load average: 0.16, 0.14,
0.20
Tasks: 463 total,   1 running, 462 sleeping,   0 stopped,   0 zombie
Cpu(s):  0.5%us,  0.2%sy,  0.0%ni, 99.3%id,  0.0%wa,  0.0%hi,
0.0%si,  0.0%st
Mem:   8310716k total,  5307148k used,  3003568k free,   341012k
buffers
Swap:  8193140k total,      628k used,  8192512k free,  3980920k
cached

On Oct 21, 9:25 am, Cedric Fontaine <cfonta...@spidmail.net> wrote:

I have responded off-list to Cedric on this problem. The qmlnxd process that
is consuming all the processor power is looping trying to wait on a
semaphore that doesn't belong to QM at all. We are investigating the
possibility of a timing glitch that could cause a loop if QM is shutdown
while qmlnxd is scanning for "lost" processes.  So far, we have failed to
find anything.

My recommendation to Cedric was to kill the looping process as is appears to
be a remnant of a previous QM startup. We are still waiting confirmation
from Cedric about whether this is a one off occurrence or a repeatable
situation. This will help us diagnose the issue.

We have identified the probable cause of this problem. The fix will be in
release 2.9-10 though no date is set for this yet. If you need an urgent
pre-build, please let me know.