Re: Porting GNOME to Wayland NFS performance in GNOME 3
On Tue, Mar 19, 2013 at 12:48:40AM +0100, stefan skoglund(agj) wrote: The RedHat thing is a really longlived bug in redhats bugzilla about gvfs metadata induced overload of NFS servers. That bug is rather bad and i think that if it isn't resolved it will make GNOME3 impossible to run in NFS-environments. Could you please stop using vague phrases like 'RedHat thing'. I have no idea what you're talking about. Either be specific, or just stop. It seems you're referring to GNOME, which is not a 'RedHat thing'. But actually I have no clue at all what you're suggesting with that. I am pretty sure that you're wrong though. -- Regards, Olav ___ desktop-devel-list mailing list desktop-devel-list@gnome.org https://mail.gnome.org/mailman/listinfo/desktop-devel-list
Re: Porting GNOME to Wayland NFS performance in GNOME 3
We have a gnome-integration list dedicated to integrating GNOME into environments. That would be a great place to discuss and figure it out. I'd like to see if we can make GNOME better in environments like yours. Login performance is slow even without NFS. Boot up performance to GDM seems to work pretty good. But after that, it has sucked ass. I refer you to this; https://mail.gnome.org/archives/gnome-shell-list/2012-May/msg00089.html on trying to find some real values on debugging the slow start up. I fear though that is out of topic for this mailing list. So follow ups to gnome-integration would be appreciated. sri On Mon, Mar 18, 2013 at 4:48 PM, stefan skoglund(agj) stefan.skogl...@agj.net wrote: mån 2013-03-18 klockan 09:10 -0700 skrev Sriram Ramkrishna: On Mon, Mar 18, 2013 at 7:58 AM, stefan skoglund(agj) stefan.skogl...@agj.net wrote: fre 2013-03-15 klockan 14:32 -0400 skrev Matthias Clasen: I dont think Redhat wants to have the same type of conversation they had with an client about GVFS bad behaviour when running over NFS if an wayland compositor is sensitive to the same type of race condition as gvfsd. In general, using NFS is a bad idea for a desktop in any case. As you say there is any number of conditions due to locking that could cause race conditions. OR is the gnome community of the belief that NFS-accessed home directories is obsolete ? The race condition in gvfsd can be triggered in the use case of a single user desktop on a single machine but said machine needs to be heavily loaded. Speaking of someone who has been in a very large enterprise environment where our home directories were all NFS mounted, we never ran into these issues. Why? Because we all ran fvwm and not a full blown desktop OS. I have a university lab setup with gnome 3.6 desktop environment in debian wheezy and Kerberized NFS-access to the home directory (the server is a Nexenta Appliance.) It is enough to say that login performance is abysmal. I think this steems from the heavy usage of dconf at login-time (at least 1 minute from login in gdm to a working desktop.) This is on 4 year old HP AMD64 hardware and intel i745 (?) hardware. I occasionally also have a bit of trouble with Pulseaudio's .pulse directory in this environment. A pristine KDE in the same setup has very nice login performance so do enlightenment (of course.) The RedHat thing is a really longlived bug in redhats bugzilla about gvfs metadata induced overload of NFS servers. That bug is rather bad and i think that if it isn't resolved it will make GNOME3 impossible to run in NFS-environments. I hope that Weston (for example) doesn't create a situation like that but i'm pessimistic. ___ desktop-devel-list mailing list desktop-devel-list@gnome.org https://mail.gnome.org/mailman/listinfo/desktop-devel-list
Re: Porting GNOME to Wayland NFS performance in GNOME 3
mån 2013-03-18 klockan 09:10 -0700 skrev Sriram Ramkrishna: On Mon, Mar 18, 2013 at 7:58 AM, stefan skoglund(agj) stefan.skogl...@agj.net wrote: fre 2013-03-15 klockan 14:32 -0400 skrev Matthias Clasen: I dont think Redhat wants to have the same type of conversation they had with an client about GVFS bad behaviour when running over NFS if an wayland compositor is sensitive to the same type of race condition as gvfsd. In general, using NFS is a bad idea for a desktop in any case. As you say there is any number of conditions due to locking that could cause race conditions. OR is the gnome community of the belief that NFS-accessed home directories is obsolete ? The race condition in gvfsd can be triggered in the use case of a single user desktop on a single machine but said machine needs to be heavily loaded. Speaking of someone who has been in a very large enterprise environment where our home directories were all NFS mounted, we never ran into these issues. Why? Because we all ran fvwm and not a full blown desktop OS. I have a university lab setup with gnome 3.6 desktop environment in debian wheezy and Kerberized NFS-access to the home directory (the server is a Nexenta Appliance.) It is enough to say that login performance is abysmal. I think this steems from the heavy usage of dconf at login-time (at least 1 minute from login in gdm to a working desktop.) This is on 4 year old HP AMD64 hardware and intel i745 (?) hardware. I occasionally also have a bit of trouble with Pulseaudio's .pulse directory in this environment. A pristine KDE in the same setup has very nice login performance so do enlightenment (of course.) The RedHat thing is a really longlived bug in redhats bugzilla about gvfs metadata induced overload of NFS servers. That bug is rather bad and i think that if it isn't resolved it will make GNOME3 impossible to run in NFS-environments. I hope that Weston (for example) doesn't create a situation like that but i'm pessimistic. ___ desktop-devel-list mailing list desktop-devel-list@gnome.org https://mail.gnome.org/mailman/listinfo/desktop-devel-list
