| iGadget | hi | 11:35 |
|---|---|---|
| iGadget | I am experienced abnormal high load on 12.04 beta2 when my systems are completely idle. Which package should I use to report this issue on? | 11:37 |
| iGadget | experiencing | 11:37 |
| phillw | chilicui1: you've lost your cloak? | 16:29 |
| chilicui1 | phillw: yep, connection troubles >_< | 16:30 |
| phillw | hmmm, I know that feeling. | 16:30 |
| iGadget | hi | 17:48 |
| iGadget | I am experiencing abnormal high load on 12.04 beta2 when my systems are completely idle. Which package should I use to report this issue on? | 17:48 |
| kanliot | igadget can you be specific on what is causing the load? | 18:15 |
| iGadget | kanliot, well I've been trying to find out | 18:19 |
| iGadget | htop shows several processes taking very few percentage of cpu now and then | 18:19 |
| iGadget | but the load of the machines is constantly between 0.5 and 1.5 | 18:19 |
| kanliot | .5 %? | 18:20 |
| kanliot | you can also see total cpu time | 18:20 |
| kanliot | like seconds of cpu time for each process | 18:20 |
| iGadget | ok... | 18:20 |
| iGadget | well in that case it's unit-panel-service and compiz | 18:21 |
| iGadget | ...and empathy-chat | 18:21 |
| iGadget | but on the other systems I'm not using the latter | 18:21 |
| iGadget | oh and hud-service as well, although less | 18:22 |
| kanliot | really the devs need numbers | 18:22 |
| iGadget | so how can I obtain those numbers | 18:22 |
| kanliot | like empathy-chat was using 200 seconds of cpu before | 18:22 |
| iGadget | ? | 18:22 |
| kanliot | and now empathy-chat is using 500 seconds | 18:22 |
| iGadget | so... I have to revert back to 11.10 | 18:23 |
| iGadget | check the numbers there | 18:23 |
| kanliot | yeah that would help | 18:23 |
| kanliot | i know some people do power regression testing | 18:23 |
| kanliot | like phoronix | 18:24 |
| kanliot | they test the power used at idle | 18:24 |
| kanliot | for the new ubuntu versions | 18:24 |
| iGadget | yeah I read about those tests | 18:24 |
| kanliot | so you can check their results when they come out | 18:24 |
| iGadget | but still... an idle system should not have a load average between 0.5 and 1.5 | 18:24 |
| iGadget | that's just absurd | 18:24 |
| kanliot | %? | 18:25 |
| kanliot | .5 % | 18:25 |
| kanliot | ? | 18:25 |
| iGadget | it's the numbers behind 'Load average:' in htop | 18:25 |
| iGadget | right now on this system - 0.81, 0.82, 0.83 | 18:25 |
| iGadget | I've been told that 1.0 is one cpu being used 100% | 18:26 |
| iGadget | so 0.5 would equal 50% usage of 1 core, right? | 18:26 |
| kanliot | yeah | 18:28 |
| kanliot | actually i just read about load averages just now | 18:28 |
| kanliot | full disclosure | 18:28 |
| kanliot | something has to be eating your cpu | 18:29 |
| iGadget | indeed... and it's present on all 3 systems I'm running 12.04 on | 18:29 |
| kanliot | you can sort by cpu | 18:30 |
| iGadget | cpu% you mean? | 18:30 |
| kanliot | yeah | 18:30 |
| iGadget | that's default, yes | 18:30 |
| kanliot | f6 | 18:30 |
| iGadget | and what I'm seeing there is the processes I mentioned before | 18:30 |
| kanliot | and they are way below 25%? | 18:31 |
| kanliot | unit-panel-service and compiz | 18:31 |
| kanliot | ...and empathy-chat | 18:31 |
| iGadget | and there are a few more which take 1 to 2% each | 18:32 |
| kanliot | what's the highest percentage? | 18:32 |
| iGadget | Is there any way to record htop's output? | 18:32 |
| kanliot | and how many cores on that pc? | 18:32 |
| kanliot | i donno | 18:32 |
| iGadget | highest varies all the time | 18:33 |
| iGadget | one moment it's compiz | 18:33 |
| iGadget | then it's unit-panel-service | 18:33 |
| kanliot | how much cpu time on unit-panelservice | 18:33 |
| kanliot | ? | 18:33 |
| iGadget | 2 cores with hyperthreading here, Core i5 | 18:33 |
| iGadget | varies between 2 and 3% | 18:34 |
| iGadget | sometimes higher | 18:34 |
| kanliot | well i'm stumped | 18:34 |
| iGadget | and that's the weird part I guess... the number don't add up | 18:35 |
| kanliot | i don't think kernel times are included | 18:35 |
| iGadget | I guess I can make a screencast video of htop while logged in via ssh, so the recording itself doesn't affect htop | 18:37 |
| iGadget | would that help? | 18:37 |
| kanliot | quit htop and load top | 18:38 |
| kanliot | it shows kernel times | 18:38 |
| kanliot | please | 18:38 |
| iGadget | sure... so what should I look for? | 18:39 |
| kanliot | ): 18.3%us, 2.5%sy, 0.0%ni, 78.4%id, 0.8%wa, 0.0%hi, 0.0%si, 0.0%st | 18:39 |
| kanliot | look for like 40%sy | 18:39 |
| iGadget | I'm not sure what those numbers mean | 18:39 |
| iGadget | 1.3%sy | 18:39 |
| iGadget | but load average is > 0.8 | 18:39 |
| kanliot | the rest of the line? | 18:40 |
| iGadget | it's changing like crazy, so it's just a snapshot | 18:40 |
| iGadget | Cpu(s): 6.7%us, 2.1%sy, 0.0%ni, 91.2%id, 0.0%wa, 0.0%hi, 0.0%si, 0.0%st | 18:41 |
| kanliot | 2c. CPU States | 18:41 |
| kanliot | The CPU states are shown in the Summary Area. They are always shown as a percentage and are for the time between now and the last refresh. | 18:41 |
| kanliot | us -- User CPU time | 18:41 |
| kanliot | The time the CPU has spent running users' processes that are not niced. | 18:41 |
| kanliot | sy -- System CPU time | 18:41 |
| kanliot | The time the CPU has spent running the kernel and its processes. | 18:41 |
| kanliot | ni -- Nice CPU time | 18:41 |
| kanliot | The time the CPU has spent running users' proccess that have been niced. | 18:41 |
| kanliot | wa -- iowait | 18:41 |
| kanliot | Amount of time the CPU has been waiting for I/O to complete. | 18:41 |
| kanliot | hi -- Hardware IRQ | 18:41 |
| kanliot | The amount of time the CPU has been servicing hardware interrupts. | 18:41 |
| kanliot | si -- Software Interrupts | 18:41 |
| kanliot | The amount of time the CPU has been servicing software interrupts. | 18:41 |
| kanliot | st -- Steal Time | 18:41 |
| kanliot | The amount of CPU 'stolen' from this virtual machine by the hypervisor for other tasks (such as running another virtual machine). | 18:41 |
| kanliot | at 9% cpu usage i think you should have a load of 9%*4 = .36 | 18:43 |
| kanliot | actually disregard what i said | 18:44 |
| kanliot | i donno | 18:44 |
| iGadget | I guess I will just create a screencast video and post it in a bugreport. Problem still is - what package to file the bug on? :) | 18:44 |
| kanliot | i think it's a kernel problem | 18:44 |
| iGadget | linux-image then | 18:45 |
| kanliot | ya | 18:45 |
| kanliot | good luck | 18:45 |
| iGadget | thanks | 18:46 |
| kanliot | and it would help if you gave hard numbers for load | 18:46 |
| kanliot | and kernel versions also | 18:46 |
| iGadget | well those will all be included in the video and the apport report I guess :) | 19:15 |
| iGadget | kanliot, I think I found the cause of the problem | 19:39 |
| iGadget | and it's not what I thought is was :) | 19:39 |
| kanliot | ok | 19:40 |
| iGadget | I'm using indicator-multiload as a replacement for the good 'ol gnome2 system monitor in the upper bar | 19:40 |
| iGadget | but in 12.04, this indicator causes all the beforementioned processes to start consuming cpu | 19:40 |
| iGadget | when I quit the indicator, the system load goes back to normal... | 19:41 |
| iGadget | ...while the whole point to use the indicator was to get a sens of what the overal load of my system was like :P | 19:41 |
| kanliot | yeah | 19:41 |
| kanliot | pITA | 19:41 |
| iGadget | so I'll be filing a bug alright... on indicator-multiload ;) | 19:42 |
| kanliot | i suppose you should document it | 19:42 |
| kanliot | but i donno where | 19:42 |
| kanliot | :) | 19:42 |
| === SaMe is now known as SergioMeneses1 | ||
| === SergioMeneses1 is now known as SergioMeneses | ||
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!