Hi all, Anyone running Xen 3.1 (kernel 2.6.20) on an Intel Core Duo 2 (X5355) with dom0 and xen compiled as 32bit? I''m getting random postgres and mysql database crashes on all my Core Duo 2 systems, as well as SSL errors due to memory corruption. I tested all the ram and it checks out as no errors. If anyone has had similar issues did switching to 64bit fix it? thanks, Norbert _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
Norbert Phillipps wrote:> > Anyone running Xen 3.1 (kernel 2.6.20) on an Intel Core Duo 2 (X5355) > with dom0 and xen compiled as 32bit? > > I''m getting random postgres and mysql database crashes on all my Core > Duo 2 systems, as well as SSL errors due to memory corruption. I tested > all the ram and it checks out as no errors. > > If anyone has had similar issues did switching to 64bit fix it? >I was running a 2.6.20 kernel on a Core 2 Duo system (E6320), 2GB RAM. The kernel, custom compiled under Debian Etch, used the Fedora Xen patches. I am no longer running the 2.6.20 kernel because I too experienced seemingly "random" crashes due to what appeared to be memory corruption. For me the kernel would crash. It was encountered somewhere that there was suspicion the 2.6.20 kernel (regardless of Xen patches) may have been the source for these problems, so it was discarded. 2.6.18 worked, and just today I compiled a 2.6.22 kernel (based on the Ubuntu patches.. not sure if they are also shamelessly stolen from Fedora).. but so far so good. -Matthew -- Matthew Haas Visiting Instructor Corning Community College Computer & Information Science http://lab46.corning-cc.edu/haas/home/ "Writing should be like breathing; It is one of those important things we do." -- me _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
Have you had any trouble with Ubuntu''s 2.6.22? I tried it a few weeks back and my guests (especially windows) were extremely unstable. Never had one running for more than 2 hours. Which is why I went back to 2.6.20 --Norbert On Oct 1, 2007, at 3:39 PM, Matthew Haas wrote:> Norbert Phillipps wrote: >> >> Anyone running Xen 3.1 (kernel 2.6.20) on an Intel Core Duo 2 >> (X5355) with dom0 and xen compiled as 32bit? >> I''m getting random postgres and mysql database crashes on all my >> Core Duo 2 systems, as well as SSL errors due to memory >> corruption. I tested all the ram and it checks out as no errors. >> If anyone has had similar issues did switching to 64bit fix it? >> > > I was running a 2.6.20 kernel on a Core 2 Duo system (E6320), 2GB > RAM. The kernel, custom compiled under Debian Etch, used the Fedora > Xen patches. > > I am no longer running the 2.6.20 kernel because I too experienced > seemingly "random" crashes due to what appeared to be memory > corruption. For me the kernel would crash. > > It was encountered somewhere that there was suspicion the 2.6.20 > kernel (regardless of Xen patches) may have been the source for > these problems, so it was discarded. > > 2.6.18 worked, and just today I compiled a 2.6.22 kernel (based on > the Ubuntu patches.. not sure if they are also shamelessly stolen > from Fedora).. but so far so good. > > -Matthew > -- > Matthew Haas > Visiting Instructor > Corning Community College > Computer & Information Science > http://lab46.corning-cc.edu/haas/home/ > > "Writing should be like breathing; > It is one of those important things we do." -- me > > _______________________________________________ > Xen-users mailing list > Xen-users@lists.xensource.com > http://lists.xensource.com/xen-users_______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
Norbert Phillipps wrote:> Have you had any trouble with Ubuntu''s 2.6.22? I tried it a few weeks > back and my guests (especially windows) were extremely unstable. Never > had one running for more than 2 hours. Which is why I went back to 2.6.20 >Well, my initial attempts at using it find a newly created paravirtual domU (debootstrap''ed Debian etch) seems to currently be suffering from a lack of PCI devices, as made apparent by the message: PCI: Fatal: No config space access function found Judging by a few google searches.. it would appear that I''m not the first one to have this problem, although I haven''t yet found a solution.. I must be missing some kernel option. If anyone has any suggestions, I''d be welcome to them... I''ve attached a (gzipped) copy of my kernel configuration. I *thought* I had selected everything for combined dom0/domU operation (also note- this is my 2nd attempted kernel for domU on 2.6.22-- the first one had a PCI access mode of "ANY" which I changed in the attached config to "Xen PCI Frontend"). Thanks. -Matthew -- Matthew Haas Visiting Instructor Corning Community College Computer & Information Science http://lab46.corning-cc.edu/haas/home/ "Writing should be like breathing; It is one of those important things we do." -- me _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
On Mon, Oct 01, 2007 at 10:41:05PM -0400, Matthew Haas wrote:> > PCI: Fatal: No config space access function found > > Judging by a few google searches.. it would appear that I''m not the > first one to have this problem, although I haven''t yet found a > solution.. I must be missing some kernel option.You people nailed it down to core2duo, reminds me of the tons of bugs that have been discovered for it (hey, mostly hit you under rare conditions!). As the default-way to fix those (not possible for all) getting latest bios maybe including microcode-updates shouldnt hurt. There is also a userspace-program for linux for updating the microcode, maybe that doesnt work properly under xen - and running it under pure linux wont help as the microcode is lost over reboots (dunno about kexec). Christian _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
On Oct 1, 2007, at 10:41 PM, Matthew Haas wrote:> Well, my initial attempts at using it find a newly created > paravirtual domU (debootstrap''ed Debian etch) seems to currently be > suffering from a lack of PCI devices, as made apparent by the message: > > PCI: Fatal: No config space access function found > > Judging by a few google searches.. it would appear that I''m not > the first one to have this problem, although I haven''t yet found a > solution.. I must be missing some kernel option. > > If anyone has any suggestions, I''d be welcome to them... I''ve > attached a (gzipped) copy of my kernel configuration. I *thought* I > had selected everything for combined dom0/domU operation (also > note- this is my 2nd attempted kernel for domU on 2.6.22-- the > first one had a PCI access mode of "ANY" which I changed in the > attached config to "Xen PCI Frontend").here''s a copy of the config that I used. It booted my dom0s and domUs with no problem but as I said I never got it to be stable. It''s a pretty generic config for a SATA or SAS intel box --Norbert  _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
Norbert Phillipps wrote:> > here''s a copy of the config that I used. It booted my dom0s and domUs > with no problem but as I said I never got it to be stable. It''s a > pretty generic config for a SATA or SAS intel box >Actually, it turns out I need to retract my statement of non-functionality; after a good night''s rest and some time away from the problem, it dawned on me that, upon checking-- none of my other Paravirt VMs also have any PCI buses... and it appears to be only my new 2.6.22 Paravirt DomU that spits out more serious looking messages. Since yesterday afternoon (and finding the real source of my problem-- a typo in a config file) I''ve had networking active in my domU, and I left it at 100% CPU doing a monte carlo PI approximation for just under 6 hours. I''m on it now, and it has been up for just over 15 hours. I just set it up as an LDAP and autofs client on my network, and am now leaving it running a 4 CPU MPI job (domU is configured only for 1 CPU) to intentionally peg it at a high load (3.85 load and climbing, according to domU''s top) for hours. dom0''s xm top shows that domU''s CPU at 100%. So from my standpoint, so far this 2.6.22.7 domU (on a 2.6.22.7 dom0) appears quite usable. There are some things that I should point out may differ configuration-wise and hardware-wise from you: - although I have swap configured in the kernel, this domU currently has no active swap space. - the system I''m currently running this domU on is a Q6600 (Core 2 Quad 2.40GHz). My original post reflected the system I had first attempted the 2.6.20 kernel on (a Core 2 Duo E6320), so I don''t know if there''s some significant difference in hardware that will cause any difference in behavior. I have moved my 2.6.22.7 kernel over to that same E6320 (and also a E6550), and will be testing it out there on some existing domUs before too long. I will take a look at your included config to see if there are any differences with what I have that would give me suspicion of causing problems. Or I''ll just compile yours and give it a run on a domU. -Matthew -- Matthew Haas Visiting Instructor Corning Community College Computer & Information Science http://lab46.corning-cc.edu/haas/home/ "Writing should be like breathing; It is one of those important things we do." -- me _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users