Hi I''m running a i7-920 with 4G RAM and two SATA HDs in a softraid. Dom0 (Gentoo 2.6.32 with Xen patches from Suse) is pinned to 1 CPU and only one DomU (Gentoo 2.6.31) is active. DomU''s root partition is a 180GB ext4 file, furthermore it has it''s own physical 50GB partition for the swapfile, logs, database etc. Xen is at 3.4.2 with routed networking. Both Dom0 and DomU boot just fine, but after some hours without much load, the machine becomes unresponsive and has to be resetted. (With unresponsive I mean that you can''t connect via SSH neither Dom0 nor DomU. I don''t have physical access to the hardware.) The logs however don''t show anything odd at all. One thing I''ve noted: vmstat shows roughly 2000 context switches when the machine is under no load at all. I''m still pretty noob not (yet :-) experienced with debugging Xen issues, therefore I was hoping someone with more insight could give me a few hints as to in which direction I should research. Many thanks! host : boss release : 2.6.31-xen-r7 version : #5 SMP Thu Dec 10 14:21:35 CET 2009 machine : x86_64 nr_cpus : 4 nr_nodes : 1 cores_per_socket : 4 threads_per_core : 1 cpu_mhz : 2673 hw_caps : bfebfbff:28100800:00000000:00000340:0098e3bd:00000000:00000001:00000000 virt_caps : hvm total_memory : 8183 free_memory : 5069 node_to_cpu : node0:0-3 node_to_memory : node0:5069 xen_major : 3 xen_minor : 4 xen_extra : .2 xen_caps : xen-3.0-x86_64 xen-3.0-x86_32p hvm-3.0-x86_32 hvm-3.0-x86_32p hvm-3.0-x86_64 xen_scheduler : credit xen_pagesize : 4096 platform_params : virt_start=0xffff800000000000 xen_changeset : unavailable cc_compiler : gcc version 4.3.4 (Gentoo 4.3.4 p1.0, pie-10.1.5) cc_compile_by : root cc_compile_date : Thu Dec 10 13:11:35 CET 2009 xend_config_format : 4 -- View this message in context: http://old.nabble.com/Xenned-box-becomes-unresponsive-tp26770231p26770231.html Sent from the Xen - User mailing list archive at Nabble.com. _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
Same for me on opensuse 11.2. Had to rollback to opensuse 11.1 (kernel 2.6.27.37). I guess it is when IO load is generated from domU and the domU lives in a "raw device" such as a prtition or a lvm volume. I could not crash the machine, if domU lives in a file. An alternative is to add "mem=4G" as parameter for the xen hypervisor in grub: kernel /xen.gz mem=4G See Bug: https://bugzilla.novell.com/show_bug.cgi?id=551695 Svoop wrote:> Hi > > I''m running a i7-920 with 4G RAM and two SATA HDs in a softraid. Dom0 > (Gentoo 2.6.32 with Xen patches from Suse) is pinned to 1 CPU and only one > DomU (Gentoo 2.6.31) is active. DomU''s root partition is a 180GB ext4 file, > furthermore it has it''s own physical 50GB partition for the swapfile, logs, > database etc. Xen is at 3.4.2 with routed networking. > > Both Dom0 and DomU boot just fine, but after some hours without much load, > the machine becomes unresponsive and has to be resetted. (With unresponsive > I mean that you can''t connect via SSH neither Dom0 nor DomU. I don''t have > physical access to the hardware.) The logs however don''t show anything odd > at all. > > One thing I''ve noted: vmstat shows roughly 2000 context switches when the > machine is under no load at all. > > I''m still pretty noob not (yet :-) experienced with debugging Xen issues, > therefore I was hoping someone with more insight could give me a few hints > as to in which direction I should research. > > Many thanks! > > > > host : boss > release : 2.6.31-xen-r7 > version : #5 SMP Thu Dec 10 14:21:35 CET 2009 > machine : x86_64 > nr_cpus : 4 > nr_nodes : 1 > cores_per_socket : 4 > threads_per_core : 1 > cpu_mhz : 2673 > hw_caps : > bfebfbff:28100800:00000000:00000340:0098e3bd:00000000:00000001:00000000 > virt_caps : hvm > total_memory : 8183 > free_memory : 5069 > node_to_cpu : node0:0-3 > node_to_memory : node0:5069 > xen_major : 3 > xen_minor : 4 > xen_extra : .2 > xen_caps : xen-3.0-x86_64 xen-3.0-x86_32p hvm-3.0-x86_32 > hvm-3.0-x86_32p hvm-3.0-x86_64 > xen_scheduler : credit > xen_pagesize : 4096 > platform_params : virt_start=0xffff800000000000 > xen_changeset : unavailable > cc_compiler : gcc version 4.3.4 (Gentoo 4.3.4 p1.0, pie-10.1.5) > cc_compile_by : root > cc_compile_date : Thu Dec 10 13:11:35 CET 2009 > xend_config_format : 4 > >_______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
Florian Gleixner wrote:> > Same for me on opensuse 11.2. Had to rollback to opensuse 11.1 (kernel > 2.6.27.37). I guess it is when IO load is generated from domU and the > domU lives in a "raw device" such as a prtition or a lvm volume. >But my DomU / is a file already, the phy partition is only used for swapfile, logs and db. I''ll try the mem workaround anyway. A side note: Both Dom0 and DomU kernels are set to 1000 Hz timer. Is that okay? I''ve read somewhere that the Xen kernels default to 100 Hz. Not sure if this is relevant at all on this matter though. -- View this message in context: http://old.nabble.com/Xenned-box-becomes-unresponsive-tp26770231p26771782.html Sent from the Xen - User mailing list archive at Nabble.com. _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
Florian Gleixner wrote:> > "mem=4G" as parameter for the xen hypervisor in grub: >Do the 4G have to reflect the real memory or is it part of the workaround to always set it to 4G? I''d like to try on another box which has 8G. -- View this message in context: http://old.nabble.com/Xenned-box-becomes-unresponsive-tp26770231p26779450.html Sent from the Xen - User mailing list archive at Nabble.com. _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
It limits the machine to 4G:
mem=xxx
Set the physical RAM address limit. Any RAM appearing beyond this
physical address in the memory map will be ignored. This parameter may
be specified with a B, K, M or G suffix, representing bytes, kilobytes,
megabytes and gigabytes respectively. The default unit, if no suffix is
specified, is kilobytes.
http://tx.downloads.xensource.com/downloads/docs/user/#SECTION04130000000000000000
On 12/14/2009 04:19 PM, Svoop wrote:>
>
> Florian Gleixner wrote:
>>
>> "mem=4G" as parameter for the xen hypervisor in grub:
>>
>
> Do the 4G have to reflect the real memory or is it part of the workaround
to
> always set it to 4G? I''d like to try on another box which has 8G.
>
_______________________________________________
Xen-users mailing list
Xen-users@lists.xensource.com
http://lists.xensource.com/xen-users
On Mon, Dec 14, 2009 at 4:22 PM, flo <flo@redflo.de> wrote:> It limits the machine to 4G: > > mem=xxx > Set the physical RAM address limit. Any RAM appearing beyond this > physical address in the memory map will be ignored. This parameter may > be specified with a B, K, M or G suffix, representing bytes, kilobytes, > megabytes and gigabytes respectively. The default unit, if no suffix is > specified, is kilobytes. > > http://tx.downloads.xensource.com/downloads/docs/user/#SECTION04130000000000000000 > > On 12/14/2009 04:19 PM, Svoop wrote: >> >> >> Florian Gleixner wrote: >>> >>> "mem=4G" as parameter for the xen hypervisor in grub: >>> >> >> Do the 4G have to reflect the real memory or is it part of the workaround to >> always set it to 4G? I''d like to try on another box which has 8G. >> > > > _______________________________________________ > Xen-users mailing list > Xen-users@lists.xensource.com > http://lists.xensource.com/xen-users >I have had similar problems with openSUSE dom0 kernels newer than 2.6.29, today Jan Beulich identified a problem which first appeared in 2.6.28 but which was masked by swiotlb settings in 2.6.29, he sent me a patch which seems to have fixed the problem as my system has been running 2.6.31.7 for 3 hours under heavy load with no issues at all, previously it would crash or lockup within minutes of starting some domU''s, so you may want to try using a 11.2 or head kotd kernel from ftp://ftp.suse.com/pub/projects/kernel/kotd/ in a few days time once Jan has committed a final version of the fix. Once .31 is completely stable I will release a rebased set of minimal dom0 openSUSE patches as I''ve done before and which I hope to get added to the Gentoo portage tree and will maintain for as long as openSUSE 11.2 is maintained, I will also continue to update the 2.6.29 ebuild and patches and plan to track kotd master (currently .32) until pv_ops is fully mature and functional, up until now the most up to date reliable version was 2.6.29 but with this new patch .31 is looking very positive. Andy _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
Hi Andrew Lyon wrote:> > Once .31 is completely stable I will release a rebased set of minimal > dom0 openSUSE patches as I''ve done before and which I hope to get > added to the Gentoo portage tree and will maintain for as long as >I see you''ve just added the xen-sources-2.6.31-r9.ebuild, does it contain the patch you mention already? Big thanks for your work, I (and many more I guess) appreciate it very much! If it doesn''t find it''s way into the main tree, maybe an overlay would be an option. The Gentoo Sunrise overlay is the most popular place, however, the review process usually delays ebuild updates for several days or weeks - not desirable for xen-sources. There''s also a Xen Overlay which hasn''t seen updates for 7 months. Shouldn''t be difficult to convince the overlay owner to give you write privis. http://overlays.gentoo.org/proj/xen/browser Cheers, -sven -- View this message in context: http://old.nabble.com/Xenned-box-becomes-unresponsive-tp26770231p26814936.html Sent from the Xen - User mailing list archive at Nabble.com. _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
On Wed, Dec 16, 2009 at 5:20 PM, Svoop <svoop@delirium.ch> wrote:> > Hi > > > Andrew Lyon wrote: >> >> Once .31 is completely stable I will release a rebased set of minimal >> dom0 openSUSE patches as I''ve done before and which I hope to get >> added to the Gentoo portage tree and will maintain for as long as >> > > I see you''ve just added the xen-sources-2.6.31-r9.ebuild, does it contain > the patch you mention already?Yes.> > Big thanks for your work, I (and many more I guess) appreciate it very much! > > If it doesn''t find it''s way into the main tree, maybe an overlay would be an > option. The Gentoo Sunrise overlay is the most popular place, however, the > review process usually delays ebuild updates for several days or weeks - not > desirable for xen-sources. There''s also a Xen Overlay which hasn''t seen > updates for 7 months. Shouldn''t be difficult to convince the overlay owner > to give you write privis. >I will get in touch, .31 will be maintained for a long time as it is the openSUSE 11.2 kernel and I will update it as necessary, but I also plan to release ebuilds and patches for 2.6.32 and continue to track the current stable kernel, so we should be able to have 3 new kernel builds in portage ;). Andy> http://overlays.gentoo.org/proj/xen/browser > > Cheers, -sven > -- > View this message in context: http://old.nabble.com/Xenned-box-becomes-unresponsive-tp26770231p26814936.html > Sent from the Xen - User mailing list archive at Nabble.com. > > > _______________________________________________ > 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