Re: CPU getting hot or maybe not
On 01/04/2012 08:21 AM, Adam Williamson wrote: On Tue, 2012-01-03 at 07:24 -0600, Bruno Wolff III wrote: On Tue, Jan 03, 2012 at 17:09:07 +0800, Frederic Muller wrote: On 01/03/2012 04:01 PM, drago01 wrote: This sounds like the debug enabled kernel is causing all this (there was a similar issue in the past iirc). Does running the F16 kernel on top of rawhide fix your issues? How would I do that? I've altered my setup to be able to easily switch from F16 to Rawhide while still working on my 'real work' so I don't mind giving it a go (and it's fast). Recently I haven't been running f16 kernels on rawhide because of a recent dependency change in f16 to require a higher version of a package than is in rawhide. The rawhide package has the needed change so you could do a nodeps install. The debugging kernels since 3.1 seem to be worse than they were in the past. For some people this was significantly impacting their display performance, for myself, my I/O performance tanked. I use luks on top of raid and I think the I/O is becoming CPU bound instead of disk bound on my system. My current solution is to rebuild kernels with most of the debug options turned back off. I have a git branch with change to the debugging options and when there is a new update I update my copy of the master branch, megre master to my nodebug branch and then do a local build. This is what I do too, more or less - I just run 'make release' on the kernel spec checkout - this is what the kernel maintainers use to switch between 'debug' and 'release' builds - bump the rev by .1 and add a changelog entry, then do a scratch build. My last one is here: http://koji.fedoraproject.org/koji/taskinfo?taskID=3612350 you're welcome to use it while it's still around. As Bruno says, debug kernels are much slower and also much more 'active' than release kernels; this could certainly explain the heat issue, as the laptop will rarely be able to put its CPU into sleep mode with all the debug activity going on. BTW, Frederic, your issue with getting an F17 nightly to boot is https://bugzilla.redhat.com/show_bug.cgi?id=766955 . It's well known, but no-one seems to be stepping up to fix it :( Thank you everyone for all the help. I'll look into it within a day or two hopefully. Fred -- test mailing list test@lists.fedoraproject.org To unsubscribe: https://admin.fedoraproject.org/mailman/listinfo/test
Re: CPU getting hot or maybe not
On Tue, 2012-01-03 at 07:24 -0600, Bruno Wolff III wrote: > On Tue, Jan 03, 2012 at 17:09:07 +0800, > Frederic Muller wrote: > > On 01/03/2012 04:01 PM, drago01 wrote: > > > > > >This sounds like the debug enabled kernel is causing all this (there > > >was a similar issue in the past iirc). > > >Does running the F16 kernel on top of rawhide fix your issues? > > > > How would I do that? I've altered my setup to be able to easily > > switch from F16 to Rawhide while still working on my 'real work' so > > I don't mind giving it a go (and it's fast). > > Recently I haven't been running f16 kernels on rawhide because of a > recent dependency change in f16 to require a higher version of a package > than is in rawhide. The rawhide package has the needed change so you > could do a nodeps install. > > The debugging kernels since 3.1 seem to be worse than they were in the > past. For some people this was significantly impacting their display > performance, for myself, my I/O performance tanked. I use luks on top > of raid and I think the I/O is becoming CPU bound instead of disk bound > on my system. > > My current solution is to rebuild kernels with most of the debug options > turned back off. I have a git branch with change to the debugging options > and when there is a new update I update my copy of the master branch, > megre master to my nodebug branch and then do a local build. This is what I do too, more or less - I just run 'make release' on the kernel spec checkout - this is what the kernel maintainers use to switch between 'debug' and 'release' builds - bump the rev by .1 and add a changelog entry, then do a scratch build. My last one is here: http://koji.fedoraproject.org/koji/taskinfo?taskID=3612350 you're welcome to use it while it's still around. As Bruno says, debug kernels are much slower and also much more 'active' than release kernels; this could certainly explain the heat issue, as the laptop will rarely be able to put its CPU into sleep mode with all the debug activity going on. BTW, Frederic, your issue with getting an F17 nightly to boot is https://bugzilla.redhat.com/show_bug.cgi?id=766955 . It's well known, but no-one seems to be stepping up to fix it :( -- Adam Williamson Fedora QA Community Monkey IRC: adamw | Twitter: AdamW_Fedora | identi.ca: adamwfedora http://www.happyassassin.net -- test mailing list test@lists.fedoraproject.org To unsubscribe: https://admin.fedoraproject.org/mailman/listinfo/test
Re: CPU getting hot or maybe not
On Tue, Jan 03, 2012 at 17:09:07 +0800, Frederic Muller wrote: > On 01/03/2012 04:01 PM, drago01 wrote: > > > >This sounds like the debug enabled kernel is causing all this (there > >was a similar issue in the past iirc). > >Does running the F16 kernel on top of rawhide fix your issues? > > How would I do that? I've altered my setup to be able to easily > switch from F16 to Rawhide while still working on my 'real work' so > I don't mind giving it a go (and it's fast). Recently I haven't been running f16 kernels on rawhide because of a recent dependency change in f16 to require a higher version of a package than is in rawhide. The rawhide package has the needed change so you could do a nodeps install. The debugging kernels since 3.1 seem to be worse than they were in the past. For some people this was significantly impacting their display performance, for myself, my I/O performance tanked. I use luks on top of raid and I think the I/O is becoming CPU bound instead of disk bound on my system. My current solution is to rebuild kernels with most of the debug options turned back off. I have a git branch with change to the debugging options and when there is a new update I update my copy of the master branch, megre master to my nodebug branch and then do a local build. -- test mailing list test@lists.fedoraproject.org To unsubscribe: https://admin.fedoraproject.org/mailman/listinfo/test
Re: CPU getting hot or maybe not
On 01/03/2012 04:01 PM, drago01 wrote: On Tue, Jan 3, 2012 at 7:21 AM, Frederic Muller wrote: On 01/02/2012 05:57 PM, drago01 wrote: That sounds like some process for whatever reason stresses the CPU and your hardware is broken to the point where it cannot really run at full load. I actually agree with that statement. Now on top of having a hot laptop I also have a very sluggish one under Rawhide. This sounds like the debug enabled kernel is causing all this (there was a similar issue in the past iirc). Does running the F16 kernel on top of rawhide fix your issues? How would I do that? I've altered my setup to be able to easily switch from F16 to Rawhide while still working on my 'real work' so I don't mind giving it a go (and it's fast). Thank you. Fred -- test mailing list test@lists.fedoraproject.org To unsubscribe: https://admin.fedoraproject.org/mailman/listinfo/test
Re: CPU getting hot or maybe not
On Tue, Jan 3, 2012 at 7:21 AM, Frederic Muller wrote: > On 01/02/2012 05:57 PM, drago01 wrote: >> >> That sounds like some process for whatever reason stresses the CPU and >> your hardware is broken to the point where it cannot really run at >> full load. > > > I actually agree with that statement. Now on top of having a hot laptop I > also have a very sluggish one under Rawhide. This sounds like the debug enabled kernel is causing all this (there was a similar issue in the past iirc). Does running the F16 kernel on top of rawhide fix your issues? -- test mailing list test@lists.fedoraproject.org To unsubscribe: https://admin.fedoraproject.org/mailman/listinfo/test
Re: CPU getting hot or maybe not
On 01/02/2012 05:57 PM, drago01 wrote: That sounds like some process for whatever reason stresses the CPU and your hardware is broken to the point where it cannot really run at full load. I actually agree with that statement. Now on top of having a hot laptop I also have a very sluggish one under Rawhide. Are there ways to track down this process? I am now back to F16 and do not suffer from the problem so heavily (as stated previously it only happens after hours of usage and only prevents to reboot, it doesn't shut down my machine). This is obviously a bug. Xorg is down to below 2% CPU usage FWIW. I have only looked and reported issues once the next release started its alpha period. Is it worth investigating issues under rawhide outside of the pre-release time, or should I just wait for alpha and try again? Thank you. Fred -- test mailing list test@lists.fedoraproject.org To unsubscribe: https://admin.fedoraproject.org/mailman/listinfo/test
Re: CPU getting hot or maybe not
On Mon, Jan 2, 2012 at 4:59 AM, Frederic Muller wrote: > Hi! > > As written in my previous message my laptop gets shutdown within minutes > (below 15 minutes) after using it under Rawhide. This is not something I > experienced under F16, though I have 2 add a bit of context: > - this is a T60 thinkpad (rather old piece of hardware) > - I never had fan issues under the fan broke down (or the sensor) and I > replaced it with a new one > - This happened at the same time I switch distro from Debian / Ubuntu (I was > going back and forth) to F14. > - Since my early usage of Fedora I have noticed some thermal issue > preventing me at times to restart my machine after long hours of usage or a > yum update. This had never happened before but I don't want to blame Fedora > for it at all (and rather identify the root cause of the problem). > > So now that I am with Rawhide since yesterday this thermal issue has hit me > big time making my machine unusable. TOP sometimes gives very high CPU usage > for xorg (up to 65%) but not every time. Most time xorg is at around 14% CPU > time. I have disabled my external monitor thinking it might help to ease > xorg's pain. > > Now I do touch my laptop to see if it feels hot and it is not really hot as > it could have been when I experienced the issue in the past with F14/15/16. > It could be though, that the heat didn't have enough time to propagate to > the 'outer' parts. > > How would you guys go on and troubleshoot the issue? (and yes the fan is > turning as I can heart it). > > Thanks a lot. That sounds like some process for whatever reason stresses the CPU and your hardware is broken to the point where it cannot really run at full load. Now people will start telling me "this is how laptops are" but I still call BS ... if you can't run it at full speed and still under warranty contact the vendor and demand a fix/replacement. Otherwise try to remove dust and or check your heatsink / cooling. -- test mailing list test@lists.fedoraproject.org To unsubscribe: https://admin.fedoraproject.org/mailman/listinfo/test
CPU getting hot or maybe not
Hi! As written in my previous message my laptop gets shutdown within minutes (below 15 minutes) after using it under Rawhide. This is not something I experienced under F16, though I have 2 add a bit of context: - this is a T60 thinkpad (rather old piece of hardware) - I never had fan issues under the fan broke down (or the sensor) and I replaced it with a new one - This happened at the same time I switch distro from Debian / Ubuntu (I was going back and forth) to F14. - Since my early usage of Fedora I have noticed some thermal issue preventing me at times to restart my machine after long hours of usage or a yum update. This had never happened before but I don't want to blame Fedora for it at all (and rather identify the root cause of the problem). So now that I am with Rawhide since yesterday this thermal issue has hit me big time making my machine unusable. TOP sometimes gives very high CPU usage for xorg (up to 65%) but not every time. Most time xorg is at around 14% CPU time. I have disabled my external monitor thinking it might help to ease xorg's pain. Now I do touch my laptop to see if it feels hot and it is not really hot as it could have been when I experienced the issue in the past with F14/15/16. It could be though, that the heat didn't have enough time to propagate to the 'outer' parts. How would you guys go on and troubleshoot the issue? (and yes the fan is turning as I can heart it). Thanks a lot. Fred -- test mailing list test@lists.fedoraproject.org To unsubscribe: https://admin.fedoraproject.org/mailman/listinfo/test
