Created attachment 6761 [details] lshw Running Distributed.net dnetc program gives a non usable machine. I'm running the distributed client dnetc on all my computers with no trouble since 1997. With the new mageia5 setup, the machine completely unresponsive with running it. I'm actually running it on a mageia4 install with no problem. lshw is attached dnetc is launched as a service and installed with the script below : #! /bin/bash # Installation de dnetc en tant que service sur linux-systemd # #installation de dnetc en tant que service linux system V ./dnetc -install # # creation du fichier .service echo "[Unit]" > ./dnetc.service echo "Description=distributed.net client startup/shutdown script" >> ./dnetc.service echo "After=syslog.target network.target" >> ./dnetc.service echo "" >> ./dnetc.service echo "[Service]" >> ./dnetc.service echo "Type=forking" >> ./dnetc.service echo "ExecStart=/etc/rc.d/init.d/dnetc start" >> ./dnetc.service echo "ExecStop=/etc/rc.d/init.d/dnetc stop" >> ./dnetc.service echo "ExecReload=/etc/rc.d/init.d/dnetc reload" >> ./dnetc.service echo "" >> ./dnetc.service echo "[Install]" >> ./dnetc.service echo "WantedBy=multi-user.target" >> ./dnetc.service # # copie du fichier service vers /etc/systemd/system # cp -f ./dnetc.service /etc/systemd/system/ # # activation du service pour systemd # systemctl enable dnetc.service # # lancement du service # systemctl start dnetc No problem running it on only one CPU. When running it at full steam with all CPU and hyper threading the system becomes unresponsive. I can move the mouse but clicking is useless. I can't even open a terminal because CTRL ALT F4 (or Fx...) don't work... An open Konsole is unusable, I just can strike CTRL ALT DEL to restart with my mageia4 wich is running like a charm.
Well, as dnetc is not provided by Mageia, there's not we can do... You could play with nice -20 & ionice -c3 Also, note that is quite a lot clearer: cat > /etc/systemd/system/dnetc.service << EOF [Unit] Description=distributed.net client startup/shutdown script After=syslog.target network.target [Service] Type=forking ExecStart=/etc/rc.d/init.d/dnetc start ExecStop=/etc/rc.d/init.d/dnetc stop ExecReload=/etc/rc.d/init.d/dnetc reload [Install] WantedBy=multi-user.target EOF This is not the right way to have a systemd unit if you just wrap the init script... IMHO it's better to just use the later then...
CC: (none) => mageia, thierry.vignaudComponent: Backports => RPM PackagesSource RPM: (none) => dnetc
I've upgraded a new machine from Mageia4 to Mageia5. The hardware is different : ASUS M5A88-M EVO CHIPSET 880G AMD Phenom⢠II X4 965 Black Edition (125W) 4 Gb DDR3 It's an upgrade not a new installation. No problem running the Dnetc client at full steam with the new Mageai5; :=) I'll try upgrading different kind of machine running the Dnetc client and report here the results.
The new 4.1.8 kernel resolved the problem ! :=) The failing machine is now running perfectly with Mageia5. Good job from kernel team. Thanks.
Status: NEW => RESOLVEDResolution: (none) => FIXED
Still some glitches with the new kernel. The machine becomes unresponsive during recurring tiny periods (0,5 to 1 second). I've tried this way during 10 hours but i found my machine stalled again : black screen and no possibility to cold reboot --> reset button... No problem at all with another intensive computing program : Mersenne prime. I'll try to upgrade my Mageia4 setup to Mageia5 and test with the old 3.14.43-desktop-1.mga4 kernel.
Status: RESOLVED => REOPENEDResolution: FIXED => (none)
We do usually close bug reports about issues with 3rd party software as invalid, but I'll leave this one open since tv decided to leave it open and CC coling instead. That must have been for a good reason. @ Paul is the "all CPU at 100%" part from the summary still correct?
CC: (none) => marja11Summary: mageia5 not usable with all CPU at 100% => mageia5 not usable with all CPU at 100% when using dnetc (3rd party software)
(In reply to Marja van Waes from comment #5) > We do usually close bug reports about issues with 3rd party software as > invalid, but I'll leave this one open since tv decided to leave it open and > CC coling instead. That must have been for a good reason. > > @ Paul > > is the "all CPU at 100%" part from the summary still correct? I think it's related to the power consumption options of the kernel. I've upgraded my setup from mageia4 to mageia5. When running with the 4.1.8 kernel and the dnetc program, my Belkin power plug indicates 136 W and htop shows that only 6 threads from 12 are running at 100% and the system is not fully usable. When running the 3.14.43-desktop-1.mga4 kernel in mageia5, the Belkin indicates 141 W and htop shows that the 12 threads are running at 100% CPU and the system is perfectly stable and usable. Perhaps the kernel from mageia5 is trying to moderate the power consumption too much ? or using new option(s) ? I'll try to diff config-3.14.43-desktop-1.mga4 and config-4.1.8-desktop-1.mga5 but I'm not a kernel tweaker.
(In reply to Paul Ãric Despretz from comment #6) > (In reply to Marja van Waes from comment #5) > > We do usually close bug reports about issues with 3rd party software as > > invalid, but I'll leave this one open since tv decided to leave it open and > > CC coling instead. That must have been for a good reason. > > > > @ Paul > > > > is the "all CPU at 100%" part from the summary still correct? > > I think it's related to the power consumption options of the kernel. > I've upgraded my setup from mageia4 to mageia5. > When running with the 4.1.8 kernel and the dnetc program, my Belkin power > plug indicates 136 W and htop shows that only 6 threads from 12 are running > at 100% and the system is not fully usable. > When running the 3.14.43-desktop-1.mga4 kernel in mageia5, the Belkin > indicates 141 W and htop shows that the 12 threads are running at 100% CPU > and the system is perfectly stable and usable. > Perhaps the kernel from mageia5 is trying to moderate the power consumption > too much ? or using new option(s) ? > I'll try to diff config-3.14.43-desktop-1.mga4 and > config-4.1.8-desktop-1.mga5 but I'm not a kernel tweaker. CC'ing our kernel guru :-)
CC: (none) => tmb
Last kernel (kernel-desktop-4.4.13-1.mga5-1-1.mga5) solved this. Power consumption is back to 140W, all 12 threads at 100% CPU and system fully operational and responsive. 48 hours running this kernel with no glitches. Games with wine are also back to OK. Thanks for this release and your hard work on this.
Closing as fixed then
Status: REOPENED => RESOLVEDResolution: (none) => FIXED
The "bug" is also present in mageia6 with all the actual kernels 4.9.x. But I've found the problem. :=) It's hardware related. I've changed my power supply with a newer one supporting C6/C7 power stage. These power states are only on haswell and newer processors. That's why I've never experienced the bug on the other machines I've tested. With the new power supply, all is OK on all kernels I've tested. The bug is explained and resolved for me. :=)
Version: 5 => 6