Starting sometime today, my Red Hat Enterprise Linux 3 system started consuming too much system time, about 10%. Normally, system time is far less than that. Since not much else was running at the time, I shut down BOINC (that was running two climateprediction applications and one setiathome application. The system time went back down to normal. When I restarted BOINC, I now experience too much system time again:

14:03:11 up 42 days, 22:20, 3 users, load average: 3.27, 3.20, 3.13
125 processes: 121 sleeping, 4 running, 0 zombie, 0 stopped
CPU states: cpu user nice system irq softirq iowait idle
total 2.7% 67.4% 8.1% 0.0% 0.0% 0.0% 21.4%
cpu00 1.7% 78.1% 14.7% 0.0% 0.0% 0.1% 5.1%
cpu01 5.3% 76.8% 11.2% 0.0% 0.0% 0.0% 6.5%
cpu02 1.3% 74.2% 3.0% 0.1% 0.0% 0.0% 21.2%
cpu03 2.6% 40.6% 3.6% 0.1% 0.0% 0.0% 52.8%
Mem: 4100860k av, 3039620k used, 1061240k free, 0k shrd, 150568k buff
1330776k actv, 1202088k in_d, 73804k in_c
Swap: 8193076k av, 860k used, 8192216k free 2447796k cached

PID PPID USER PRI NI SIZE RSS SHARE STAT %CPU %MEM CTIME COMMAND
27347 27344 boinc 39 19 15060 14M 1172 R N 25.0 0.3 10:26 setiathome_4.02
27348 27346 boinc 39 19 41156 40M 4068 R N 24.6 1.0 10:30 hadsm3um_4.10_i
27349 27345 boinc 39 19 66344 64M 4068 R N 24.5 1.6 10:29 hadsm3um_4.10_i
27344 1 boinc 15 0 1812 1812 1164 S 0.0 0.0 0:00 boinc_4.19_i686
27345 27344 boinc 34 19 820 820 736 S N 0.0 0.0 0:00 hadsm3_4.10_i68
27346 27344 boinc 34 19 820 820 736 S N 0.0 0.0 0:00 hadsm3_4.10_i68

Is this a bug or a new feature? I do not wish to waste so much system time.

P.S. it seems to be running hadsm3um_4.10 instead of 4.04 or whatever it used to do.


P.S.2: Your system clock seems to be about 1/2 hour off. It should be about 19:07.

---

Hi, Jean-David,

4.10 seems to do bad things to Linux machines. In the Sulfur Alpha, it caused significant slow-down from 4.09. We hoped it wouldn't carry into the public project. It seems it has. Pity.

Alpha version 4.09 was unstable and Tolu used an older version of the compiler for 4.10, at least for Alpha -- and, seemingly, for here as well.

From one of my posts on the Alpha BB, re. this new "System anchor" for my four Linux boxes:

The pattern of "System" usage is consistent across the four machines: ___/'___/'___/'___/'_... (Edit: Back-slashes for "downslope" were dropped by the BB.)
The floor is ~3.5% (5.0% for Ebox), seldom below and not observed below 2.5% (and that is infrequent). Gradual climb to peak, then gradual slide to floor. Peak is 10-11%, occasionally to 12% for the first three machines, to 14+% for Ebox.

Task Manager shows nothing similar in Windoze Sulfur 4.09 version and I've never observed more than the occasional, momentary, 0.25% System use for any other Linux CPDN version.

Configuration of the machines can be posted if you want that. (They're running SuSE 9.0/9.1.)

(I was surprised when 4.10 was released here... [I'm still running 4.04 for public-version Models].)

Jim

---

"We have met the enemy and he is us." -- Pogo
Greetings from coastal Washington state, the scenic US Pacific Northwest.

> (I was surprised when 4.10 was released here... [I'm still running 4.04 for
> public-version Models].)
>

I think it was released so it would work with BOINC version 4.2x that needs digitally signed files. Tolu responded to my question about 4.10 <a href="http://climateapps2.oucs.ox.ac.uk/cpdnboinc/forum_thread.php?id=2065">here</a>. I think it shouldn't have the Linux performance penalty we've seen in 4.10 in sulphur alpha.

&gt; &gt; (I was surprised when 4.10 was released here... [I'm still running 4.04
&gt; for
&gt; &gt; public-version Models].)
&gt; &gt;
&gt;
&gt; I think it was released so it would work with BOINC version 4.2x that needs
&gt; digitally signed files. Tolu responded to my question about 4.10 <a> href="http://climateapps2.oucs.ox.ac.uk/cpdnboinc/forum_thread.php?id=2065"&gt;here</a>.
&gt; I think it shouldn't have the Linux performance penalty we've seen in 4.10 in
&gt; sulphur alpha.
&gt;
So what should I do? Wait it out?

BTW: now one of my applications is a zombie:

PID PPID USER PRI NI SIZE RSS SHARE STAT %CPU %MEM CTIME COMMAND
6162 27344 boinc 39 19 16468 16M 1296 R N 25.0 0.4 167:39 setiathome_4.02
9884 9878 boinc 39 19 68828 67M 4156 R N 24.9 1.6 24:31 hadsm3um_4.10_i
27344 1 boinc 15 0 2060 2060 1220 S 0.0 0.0 1112m boinc_4.19_i686
9831 27344 boinc 34 19 820 820 736 S N 0.0 0.0 0:00 hadsm3_4.10_i68
9832 9831 boinc 39 19 0 0 0 Z N 0.0 0.0 0:14 hadsm3um_4.10_i
9878 27344 boinc 34 19 832 832 748 S N 0.0 0.0 0:00 hadsm3_4.10_i68

Do I wait that out too, or should I kill 9831?

P.S.: Now both are zombies:

PID PPID USER PRI NI SIZE RSS SHARE STAT %CPU %MEM CTIME COMMAND
6162 27344 boinc 39 19 15440 15M 1296 R N 24.9 0.3 206:18 setiathome_4.02
27344 1 boinc 15 0 2060 2060 1220 S 0.0 0.0 1112m boinc_4.19_i686
9831 27344 boinc 34 19 820 820 736 S N 0.0 0.0 0:00 hadsm3_4.10_i68
9832 9831 boinc 39 19 0 0 0 Z N 0.0 0.0 0:14 hadsm3um_4.10_i
9878 27344 boinc 34 19 832 832 748 S N 0.0 0.0 25:29 hadsm3_4.10_i68
10890 9878 boinc 39 19 0 0 0 Z N 0.0 0.0 24:00 hadsm3um_4.10_i

---

&gt; &gt; (I was surprised when 4.10 was released here... [I'm still running 4.04
&gt; for
&gt; &gt; public-version Models].)
&gt; &gt;
&gt;
&gt; I think it was released so it would work with BOINC version 4.2x that needs
&gt; digitally signed files. Tolu responded to my question about 4.10 <a> href="http://climateapps2.oucs.ox.ac.uk/cpdnboinc/forum_thread.php?id=2065"&gt;here</a>.
&gt; I think it shouldn't have the Linux performance penalty we've seen in 4.10 in
&gt; sulphur alpha.
&gt;

On my machine, hadsm3um_4.10_i686-pc-linux-gnu dies after a few
minutes, the model only gets to about timestamp 100. The same
machine (smp) has another 4.04 model running fine. Is there
anything I can do?

---

&gt; P.S.2: Your system clock seems to be about 1/2 hour off. It should be about 19:07.

This was discussued <a href="http://climateapps2.oucs.ox.ac.uk/cpdnboinc/forum_thread.php?id=1556"> here.</a>

Les