I have used the script iocpu.d from Chapter 27 of the Solaris Dynamic Tracing Guide to look at some of the proceses during startup of GNOME 2.12. For gnome-session I see that that the top contributors to I/O wait are nautilus 41 libORBit-2.so.0.0.0 44 gnome-panel 46 libbonoboui-2.so.0.0.0 57 libgconf-2.so.4.1.0 69 libbonobo-2.so.0.0.0 77 libgnomeui-2.so.0.1200.0 82 icon-theme.cache 169 <none> 397 Can I figure out what is contributing to <none>? This message posted from opensolaris.org
Hi Padraig,>From the Solaris Dynamic Tracing Guide''s io provider chapter:If no file information is associated with an I/O, the fi_name field will be set to the string <none>. This is frequently because the I/O had to do with file system meta data. I believe there have been some recent changes to Solaris to reduce the number of <none>s you''ll see. I don''t know of any way to explore the <none>s further. Adam On Thu, Dec 15, 2005 at 08:20:10AM -0800, Padraig O''Briain wrote:> I have used the script iocpu.d from Chapter 27 of the Solaris Dynamic Tracing Guide to look at some of the proceses during startup of GNOME 2.12. > > For gnome-session I see that that the top contributors to I/O wait are > > nautilus 41 > libORBit-2.so.0.0.0 44 > gnome-panel 46 > libbonoboui-2.so.0.0.0 57 > libgconf-2.so.4.1.0 69 > libbonobo-2.so.0.0.0 77 > libgnomeui-2.so.0.1200.0 82 > icon-theme.cache 169 > <none> 397 > > Can I figure out what is contributing to <none>? > This message posted from opensolaris.org > _______________________________________________ > dtrace-discuss mailing list > dtrace-discuss at opensolaris.org-- Adam Leventhal, Solaris Kernel Development http://blogs.sun.com/ahl
Adam Leventhal wrote:> Hi Padraig, > >>From the Solaris Dynamic Tracing Guide''s io provider chapter: > > If no file information is associated with an I/O, the fi_name field > will be set to the string <none>. > > This is frequently because the I/O had to do with file system meta data. > I believe there have been some recent changes to Solaris to reduce the > number of <none>s you''ll see. I don''t know of any way to explore the > <none>s further. > > Adam > > On Thu, Dec 15, 2005 at 08:20:10AM -0800, Padraig O''Briain wrote: >> I have used the script iocpu.d from Chapter 27 of the Solaris Dynamic Tracing Guide to look at some of the proceses during startup of GNOME 2.12. >> >> For gnome-session I see that that the top contributors to I/O wait are >> >> nautilus 41 >> libORBit-2.so.0.0.0 44 >> gnome-panel 46 >> libbonoboui-2.so.0.0.0 57 >> libgconf-2.so.4.1.0 69 >> libbonobo-2.so.0.0.0 77 >> libgnomeui-2.so.0.1200.0 82 >> icon-theme.cache 169 >> <none> 397 >> >> Can I figure out what is contributing to <none>? >> This message posted from opensolaris.org >> _______________________________________________ >> dtrace-discuss mailing list >> dtrace-discuss at opensolaris.org >I''ve seen these when the section in question is the executable; I don''t know why yet. - Bart -- Bart Smaalders Solaris Kernel Performance barts at cyber.eng.sun.com http://blogs.sun.com/barts
> >On Thu, Dec 15, 2005 at 08:20:10AM -0800, Padraig O''Briain wrote: > >>I have used the script iocpu.d from Chapter 27 of the Solaris Dynamic > >>Tracing Guide to look at some of the proceses during startup of GNOME > >>2.12. > >> > >>For gnome-session I see that that the top contributors to I/O wait are > >> > >> nautilus 41 > >> libORBit-2.so.0.0.0 44 > >> gnome-panel 46 > >> libbonoboui-2.so.0.0.0 57 > >> libgconf-2.so.4.1.0 69 > >> libbonobo-2.so.0.0.0 77 > >> libgnomeui-2.so.0.1200.0 82 > >> icon-theme.cache 169 > >> <none> 397 > >> > >>Can I figure out what is contributing to <none>? > > I''ve seen these when the section in question is the executable; > I don''t know why yet.It''s the atime update on the i-node. You can get rid of these by telling the filesystem that you don''t want to update atimes. On UFS, this is done by mounting with "noatime"; on ZFS, this is done by setting the "atime" property to "off". Note that in either case you are in the realm of POSIX non-conformance -- POSIX requires that the atime is updated whenever the file is accessed. - Bryan -------------------------------------------------------------------------- Bryan Cantrill, Solaris Kernel Development. http://blogs.sun.com/bmc
Bryan Cantrill''s email at 12/16/2005 12:32 PM, said:>>>On Thu, Dec 15, 2005 at 08:20:10AM -0800, Padraig O''Briain wrote: >>> >>>>I have used the script iocpu.d from Chapter 27 of the Solaris Dynamic >>>>Tracing Guide to look at some of the proceses during startup of GNOME >>>>2.12. >>>> >>>>For gnome-session I see that that the top contributors to I/O wait are >>>> >>>> nautilus 41 >>>> libORBit-2.so.0.0.0 44 >>>> gnome-panel 46 >>>> libbonoboui-2.so.0.0.0 57 >>>> libgconf-2.so.4.1.0 69 >>>> libbonobo-2.so.0.0.0 77 >>>> libgnomeui-2.so.0.1200.0 82 >>>> icon-theme.cache 169 >>>> <none> 397 >>>> >>>>Can I figure out what is contributing to <none>? >> >>I''ve seen these when the section in question is the executable; >>I don''t know why yet. > > > It''s the atime update on the i-node. You can get rid of these by telling > the filesystem that you don''t want to update atimes. On UFS, this is > done by mounting with "noatime"; on ZFS, this is done by setting the > "atime" property to "off". Note that in either case you are in the realm > of POSIX non-conformance -- POSIX requires that the atime is updated > whenever the file is accessed. > > - Bryan > > -------------------------------------------------------------------------- > Bryan Cantrill, Solaris Kernel Development. http://blogs.sun.com/bmc > _______________________________________________ > dtrace-discuss mailing list > dtrace-discuss at opensolaris.org > >... or when you run VxFS, you get nothing but <none>''s. I wrote a script for a customer that would report the appropriate names even for VxFS, but alas I don''t have a copy. It was trivial to do, so if I get access to VxFS anytime soon I will recreate it. Thanks, Jarod