Hi, We just migrated xen on gentoo on amd64 from 4.0 to 4.1 and encountered a few issues that were not documented. Here is how we solved them and what issues are remaining: 1. Our config files were referring to the LVM devices like /dev/vg/domU-root with disk = [ ''phy:vg/domU-root,sda1,w'' ]. This does not work anymore, we have to include ''/dev/'': disk = [ ''phy:/dev/vg/domU-root,sda1,w'' ]. 2. We were using bridges to map the 4 ethernet cards to virtual machines so that virtual machines could have their own IP addresses. In xen 4.1, it is recommended to configure bridges by ourselves and not use xend although xend would still work. We could not get the bridges to work and went on using xend. However, the network-bridge script now includes a test: if a bridge already exist, then it would do nothing. The problem with that is that our wrapper script for network-bridge would only create the first bridge. So we had to comment out the test in network-bridge, and it would work. Did anybody successfully create bridges on gentoo? Meanwhile, we found that if a domU is started and tries to access a bridge that does not exist, dom0 would stop working. Shall it be a bug? 3. I believe we have memory issues when running a windows domU. We use to allocate 512MB to domU and found that qemu would quickly use most of it and after a while, dom0 would stop working. Now we allocate 4GB, made sure there is no ballooning and it appears to work fine. Memory leaks by qemu are widely reported but I wonder why we did not have such issues before xen 4.1 update. 4. We might also have issues with CPUs. dom0 and all domU were authorized to use all CPUs, and we saw the whole system freezing often and quickly when mounting in charge after the 4.1 update. After we spinned one CPU to dom0 we haven''t seen this issue. Would this be a bug? Any hint or help on finding better solutions would be appreciated Thanks, Sylvain _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
On Sun, Apr 10, 2011 at 7:56 AM, Sylvain CHEVALIER <sylvain.chevalier@gmail.com> wrote:> Hi, > > We just migrated xen on gentoo on amd64 from 4.0 to 4.1 and > encountered a few issues that were not documented. Here is how we > solved them and what issues are remaining: > > 2. We were using bridges to map the 4 ethernet cards to virtual > machines so that virtual machines could have their own IP addresses. > In xen 4.1, it is recommended to configure bridges by ourselves and > not use xend although xend would still work. We could not get the > bridges to work and went on using xend. However, the network-bridge > script now includes a test: if a bridge already exist, then it would > do nothing. The problem with that is that our wrapper script for > network-bridge would only create the first bridge. So we had to > comment out the test in network-bridge, and it would work. Did anybody > successfully create bridges on gentoo? > Meanwhile, we found that if a domU is started and tries to access a > bridge that does not exist, dom0 would stop working. Shall it be a > bug? >I''ve also noticed this issue. The test in network-bridge is quite lousy, as there could be countless setups with multiple non-xen related bridges. However, in the brief time I''ve spent with Xen-4.1 (I''ve encountered other issues, see my email on the list), I''ve managed to manually create a bridge interface. In the /etc/conf.d/net file, replace your config_ethX (in this case, X=1) line with the following: bridge_add_eth1="xenbr0" config_eth1=( "null" ) config_xenbr0=( "172.18.0.1/24" "2001:470:1f0b:11b9::1/64" ) brctl_xenbr0=( "stp off" ) Of course, replace the numbers to suit your own setup. Also, in the /etc/xen/xend-config.sxp, make sure you specify the bridge that Xen should use, especially if you have more than one Find the line (vif-script vif-bridge) and replace it with (vif-script ''vif-bridge bridge=xenbr0'') This should tell Xen to add the vif interfaces to the xenbr0 bridge. I hope this helps you out. _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
2011/4/10 Claudiu Curcă <alexstrasza2@gmail.com>:> On Sun, Apr 10, 2011 at 7:56 AM, Sylvain CHEVALIER > <sylvain.chevalier@gmail.com> wrote: >> Hi, >> >> We just migrated xen on gentoo on amd64 from 4.0 to 4.1 and >> encountered a few issues that were not documented. Here is how we >> solved them and what issues are remaining: >> >> 2. We were using bridges to map the 4 ethernet cards to virtual >> machines so that virtual machines could have their own IP addresses. >> In xen 4.1, it is recommended to configure bridges by ourselves and >> not use xend although xend would still work. We could not get the >> bridges to work and went on using xend. However, the network-bridge >> script now includes a test: if a bridge already exist, then it would >> do nothing. The problem with that is that our wrapper script for >> network-bridge would only create the first bridge. So we had to >> comment out the test in network-bridge, and it would work. Did anybody >> successfully create bridges on gentoo? >> Meanwhile, we found that if a domU is started and tries to access a >> bridge that does not exist, dom0 would stop working. Shall it be a >> bug? >> > I''ve also noticed this issue. The test in network-bridge is quite > lousy, as there could be countless setups with multiple non-xen > related bridges. > However, in the brief time I''ve spent with Xen-4.1 (I''ve encountered > other issues, see my email on the list), I''ve managed to manually > create a bridge interface. > > In the /etc/conf.d/net file, replace your config_ethX (in this case, > X=1) line with the following: > > bridge_add_eth1="xenbr0" > config_eth1=( "null" ) > config_xenbr0=( "172.18.0.1/24" "2001:470:1f0b:11b9::1/64" ) > brctl_xenbr0=( "stp off" ) > > Of course, replace the numbers to suit your own setup. > > Also, in the /etc/xen/xend-config.sxp, make sure you specify the > bridge that Xen should use, especially if you have more than one > Find the line > > (vif-script vif-bridge) > > and replace it with > > (vif-script ''vif-bridge bridge=xenbr0'') > > This should tell Xen to add the vif interfaces to the xenbr0 bridge. > > I hope this helps you out. >Right, I almost missed this one: Don''t forget to create the net.xenbr0 symlink in /etc/init.d and also add it to the boot runlevel using rc-update. If you don''t do this, your bridge will not come up. _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
Pasi Kärkkäinen
2011-Apr-10 19:26 UTC
[Xen-devel] Re: [Xen-users] Issues migrating to xen 4.1 / lvm vg for disk
On Sun, Apr 10, 2011 at 12:56:05PM +0800, Sylvain CHEVALIER wrote:> Hi, > > We just migrated xen on gentoo on amd64 from 4.0 to 4.1 and > encountered a few issues that were not documented. Here is how we > solved them and what issues are remaining: > > 1. Our config files were referring to the LVM devices like > /dev/vg/domU-root with disk = [ ''phy:vg/domU-root,sda1,w'' ]. This does > not work anymore, we have to include ''/dev/'': disk = [ > ''phy:/dev/vg/domU-root,sda1,w'' ]. >Hello, I added xen-devel to CC list. (I''ll reply your other questions on another email.) -- Pasi _______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel
On Sun, Apr 10, 2011 at 12:56:05PM +0800, Sylvain CHEVALIER wrote:> Hi, > > We just migrated xen on gentoo on amd64 from 4.0 to 4.1 and > encountered a few issues that were not documented. Here is how we > solved them and what issues are remaining: > > 1. Our config files were referring to the LVM devices like > /dev/vg/domU-root with disk = [ ''phy:vg/domU-root,sda1,w'' ]. This does > not work anymore, we have to include ''/dev/'': disk = [ > ''phy:/dev/vg/domU-root,sda1,w'' ]. >Hello, I replied this 1. in another email.> 2. We were using bridges to map the 4 ethernet cards to virtual > machines so that virtual machines could have their own IP addresses. > In xen 4.1, it is recommended to configure bridges by ourselves and > not use xend although xend would still work. We could not get the > bridges to work and went on using xend. However, the network-bridge > script now includes a test: if a bridge already exist, then it would > do nothing. The problem with that is that our wrapper script for > network-bridge would only create the first bridge. So we had to > comment out the test in network-bridge, and it would work. Did anybody > successfully create bridges on gentoo? > Meanwhile, we found that if a domU is started and tries to access a > bridge that does not exist, dom0 would stop working. Shall it be a > bug? >What do you mean dom0 would stop working? Network dies? Does it work from the console? And yes, it''s recommended to configure networking using the distro network scripts, see here for examples: http://wiki.xen.org/xenwiki/HostConfiguration/Networking> 3. I believe we have memory issues when running a windows domU. We use > to allocate 512MB to domU and found that qemu would quickly use most > of it and after a while, dom0 would stop working. Now we allocate 4GB, > made sure there is no ballooning and it appears to work fine. Memory > leaks by qemu are widely reported but I wonder why we did not have > such issues before xen 4.1 update. >Again what do you mean with dom0 would stop working? dom0 kernel crash? hypervisor crash? And did you mean qemu-dm process in dom0 is leaking memory?> 4. We might also have issues with CPUs. dom0 and all domU were > authorized to use all CPUs, and we saw the whole system freezing often > and quickly when mounting in charge after the 4.1 update. After we > spinned one CPU to dom0 we haven''t seen this issue. Would this be a > bug? >This definitely sounds like a bug. Please set up a serial console and capture the error/crash messages: http://wiki.xensource.com/xenwiki/XenSerialConsole Also: what dom0 kernel version are you using?> Any hint or help on finding better solutions would be appreciated >-- Pasi _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
> And did you mean qemu-dm process in dom0 is leaking memory? >I believe so. After running the whole week-end, with one instance of windows 2008 used as a domain controller, qemu-dm uses 2GB of RAM. Running pmap, I am getting lots of lines like this one: 00007fcd73f2e000 1024K rw-s- /proc/xen/privcmd $ pmap 10423| grep /proc/xen/privcmd | wc -l 1827 Any clue? Sylvain _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
Many thanks Pasi and Claudiu for helping and sharing. I am sorry for not giving much more details about our experiences with dom0 hanging, but I don''t want to touch our server right now as it is running fine (except for qemu-dm memory leak) and people are using it. I hope to set up another server this week and experiment with it. These issues were similar to what Claudu described, but again, after applying the fixes I described, it has been working fine, and we did not have to revert it to 4.0. I think we had terminal issues, with any command run locally or remotely in dom0 giving an ''input/output'' error message), and we could not establish any other connection to it. We are using the latest kernel on gentoo: 2.6.34-xen-r4 and had similar issues with the previous one (2.6.34-xen-r3). Thanks, Sylvain On Mon, Apr 11, 2011 at 3:30 AM, Pasi Kärkkäinen <pasik@iki.fi> wrote:> On Sun, Apr 10, 2011 at 12:56:05PM +0800, Sylvain CHEVALIER wrote: >> Hi, >> >> We just migrated xen on gentoo on amd64 from 4.0 to 4.1 and >> encountered a few issues that were not documented. Here is how we >> solved them and what issues are remaining: >> >> 1. Our config files were referring to the LVM devices like >> /dev/vg/domU-root with disk = [ ''phy:vg/domU-root,sda1,w'' ]. This does >> not work anymore, we have to include ''/dev/'': disk = [ >> ''phy:/dev/vg/domU-root,sda1,w'' ]. >> > > Hello, > > I replied this 1. in another email. > >> 2. We were using bridges to map the 4 ethernet cards to virtual >> machines so that virtual machines could have their own IP addresses. >> In xen 4.1, it is recommended to configure bridges by ourselves and >> not use xend although xend would still work. We could not get the >> bridges to work and went on using xend. However, the network-bridge >> script now includes a test: if a bridge already exist, then it would >> do nothing. The problem with that is that our wrapper script for >> network-bridge would only create the first bridge. So we had to >> comment out the test in network-bridge, and it would work. Did anybody >> successfully create bridges on gentoo? >> Meanwhile, we found that if a domU is started and tries to access a >> bridge that does not exist, dom0 would stop working. Shall it be a >> bug? >> > > What do you mean dom0 would stop working? Network dies? > Does it work from the console? > > And yes, it''s recommended to configure networking using the distro > network scripts, see here for examples: http://wiki.xen.org/xenwiki/HostConfiguration/Networking > > >> 3. I believe we have memory issues when running a windows domU. We use >> to allocate 512MB to domU and found that qemu would quickly use most >> of it and after a while, dom0 would stop working. Now we allocate 4GB, >> made sure there is no ballooning and it appears to work fine. Memory >> leaks by qemu are widely reported but I wonder why we did not have >> such issues before xen 4.1 update. >> > > Again what do you mean with dom0 would stop working? dom0 kernel crash? hypervisor crash? > And did you mean qemu-dm process in dom0 is leaking memory? > >> 4. We might also have issues with CPUs. dom0 and all domU were >> authorized to use all CPUs, and we saw the whole system freezing often >> and quickly when mounting in charge after the 4.1 update. After we >> spinned one CPU to dom0 we haven''t seen this issue. Would this be a >> bug? >> > > This definitely sounds like a bug. > > Please set up a serial console and capture the error/crash messages: > http://wiki.xensource.com/xenwiki/XenSerialConsole > > Also: what dom0 kernel version are you using? > >> Any hint or help on finding better solutions would be appreciated >> > > -- Pasi > >_______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
Pasi Kärkkäinen
2011-Apr-11 06:54 UTC
[Xen-devel] Re: [Xen-users] Issues migrating to xen 4.1 / qemu-dm memory leak?
On Mon, Apr 11, 2011 at 08:41:59AM +0800, Sylvain CHEVALIER wrote:> > And did you mean qemu-dm process in dom0 is leaking memory? > > > > I believe so. After running the whole week-end, with one instance of > windows 2008 used as a domain controller, qemu-dm uses 2GB of RAM. > Running pmap, I am getting lots of lines like this one: > > 00007fcd73f2e000 1024K rw-s- /proc/xen/privcmd > > $ pmap 10423| grep /proc/xen/privcmd | wc -l > 1827 > > Any clue? >Ok. So, a summary: - VM has 512 MB of memory: qemu-dm leaks memory. - VM has 4GB of memory: qemu-dm doesn''t leak memory. Is that what you meant? btw. is that Windows running any PV drivers or pure HVM with emulated devices? -- Pasi _______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel
Sylvain CHEVALIER
2011-Apr-11 07:05 UTC
Re: [Xen-users] Issues migrating to xen 4.1 / qemu-dm memory leak?
> Ok. So, a summary: > > - VM has 512 MB of memory: qemu-dm leaks memory. > - VM has 4GB of memory: qemu-dm doesn''t leak memory. > > Is that what you meant?Oups, I am sorry, I am realizing that I made a typo in my original message, I am only talking about dom0, not the VM. qemu-dm is leaking memory, so if I only allocate 512MB to dom0 it will quickly run out of memory. When I allocate more (4GB), I do not run out. At least until now.> > btw. is that Windows running any PV drivers or pure HVM with emulated devices? >Pure HVM with emulated devices with qemu, I haven''t looked at PV drivers. Sylvain _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
Pasi Kärkkäinen
2011-Apr-28 19:33 UTC
Re: [Xen-devel] Re: [Xen-users] Issues migrating to xen 4.1 / lvm vg for disk
Hello, (added Ian as CC). While talking about supported disk configuration syntax, I remembered this report again.. Sylvain: Did you use ''xl'' or ''xm'' ? -- Pasi On Sun, Apr 10, 2011 at 10:26:22PM +0300, Pasi Kärkkäinen wrote:> On Sun, Apr 10, 2011 at 12:56:05PM +0800, Sylvain CHEVALIER wrote: > > Hi, > > > > We just migrated xen on gentoo on amd64 from 4.0 to 4.1 and > > encountered a few issues that were not documented. Here is how we > > solved them and what issues are remaining: > > > > 1. Our config files were referring to the LVM devices like > > /dev/vg/domU-root with disk = [ ''phy:vg/domU-root,sda1,w'' ]. This does > > not work anymore, we have to include ''/dev/'': disk = [ > > ''phy:/dev/vg/domU-root,sda1,w'' ]. > > > > Hello, > > I added xen-devel to CC list. > (I''ll reply your other questions on another email.) > > -- Pasi >_______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel
Pasi Kärkkäinen
2011-Apr-28 19:42 UTC
Re: [Xen-devel] Re: [Xen-users] Issues migrating to xen 4.1 / qemu-dm memory leak?
On Mon, Apr 11, 2011 at 09:54:27AM +0300, Pasi Kärkkäinen wrote:> On Mon, Apr 11, 2011 at 08:41:59AM +0800, Sylvain CHEVALIER wrote: > > > And did you mean qemu-dm process in dom0 is leaking memory? > > > > > > > I believe so. After running the whole week-end, with one instance of > > windows 2008 used as a domain controller, qemu-dm uses 2GB of RAM. > > Running pmap, I am getting lots of lines like this one: > > > > 00007fcd73f2e000 1024K rw-s- /proc/xen/privcmd > > > > $ pmap 10423| grep /proc/xen/privcmd | wc -l > > 1827 > > > > Any clue? > > > > Ok. So, a summary: > > - VM has 512 MB of memory: qemu-dm leaks memory. > - VM has 4GB of memory: qemu-dm doesn''t leak memory. > > Is that what you meant? > > btw. is that Windows running any PV drivers or pure HVM with emulated devices? >Hello again, Sylvain: Are you still seeing this qemu-dm memleak problem? Did you resolve it? -- Pasi _______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel
Sylvain CHEVALIER
2011-Apr-29 00:48 UTC
Re: [Xen-devel] Re: [Xen-users] Issues migrating to xen 4.1 / lvm vg for disk
Hi Pasi, Thanks a lot for following up. We had to change this part of our config files when we moved from xm to xl. Sylvain On Fri, Apr 29, 2011 at 3:33 AM, Pasi Kärkkäinen <pasik@iki.fi> wrote:> Hello, > > (added Ian as CC). > > While talking about supported disk configuration syntax, > I remembered this report again.. > > Sylvain: Did you use ''xl'' or ''xm'' ? > > -- Pasi > > On Sun, Apr 10, 2011 at 10:26:22PM +0300, Pasi Kärkkäinen wrote: >> On Sun, Apr 10, 2011 at 12:56:05PM +0800, Sylvain CHEVALIER wrote: >> > Hi, >> > >> > We just migrated xen on gentoo on amd64 from 4.0 to 4.1 and >> > encountered a few issues that were not documented. Here is how we >> > solved them and what issues are remaining: >> > >> > 1. Our config files were referring to the LVM devices like >> > /dev/vg/domU-root with disk = [ ''phy:vg/domU-root,sda1,w'' ]. This does >> > not work anymore, we have to include ''/dev/'': disk = [ >> > ''phy:/dev/vg/domU-root,sda1,w'' ]. >> > >> >> Hello, >> >> I added xen-devel to CC list. >> (I''ll reply your other questions on another email.) >> >> -- Pasi >> > >_______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel
Sylvain CHEVALIER
2011-Apr-29 00:56 UTC
Re: [Xen-devel] Re: [Xen-users] Issues migrating to xen 4.1 / qemu-dm memory leak?
Hi Pasi, I think it does leak memory although I am not completely sure. What I am seeing is that it uses almost 2GB of memory, and that it is (slowly) growing. Before we updated to xen 4.1 (which included a qemu update), we only had 512MB for dom0. Right now, we have (24576 is qemu''s pid): $ pmap 24576 total 1765848K $ pmap 24576 | grep /proc/xen/privcmd | wc -l 1624 Thanks, Sylvain On Fri, Apr 29, 2011 at 3:42 AM, Pasi Kärkkäinen <pasik@iki.fi> wrote:> On Mon, Apr 11, 2011 at 09:54:27AM +0300, Pasi Kärkkäinen wrote: >> On Mon, Apr 11, 2011 at 08:41:59AM +0800, Sylvain CHEVALIER wrote: >> > > And did you mean qemu-dm process in dom0 is leaking memory? >> > > >> > >> > I believe so. After running the whole week-end, with one instance of >> > windows 2008 used as a domain controller, qemu-dm uses 2GB of RAM. >> > Running pmap, I am getting lots of lines like this one: >> > >> > 00007fcd73f2e000 1024K rw-s- /proc/xen/privcmd >> > >> > $ pmap 10423| grep /proc/xen/privcmd | wc -l >> > 1827 >> > >> > Any clue? >> > >> >> Ok. So, a summary: >> >> - VM has 512 MB of memory: qemu-dm leaks memory. >> - VM has 4GB of memory: qemu-dm doesn''t leak memory. >> >> Is that what you meant? >> >> btw. is that Windows running any PV drivers or pure HVM with emulated devices? >> > > Hello again, > > Sylvain: Are you still seeing this qemu-dm memleak problem? Did you resolve it? > > -- Pasi > >_______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel
Ian Campbell
2011-May-03 14:13 UTC
Re: [Xen-devel] Re: [Xen-users] Issues migrating to xen 4.1 / lvm vg for disk
> >> On Sun, Apr 10, 2011 at 12:56:05PM +0800, Sylvain CHEVALIER wrote: > >> > 1. Our config files were referring to the LVM devices like > >> > /dev/vg/domU-root with disk = [ ''phy:vg/domU-root,sda1,w'' ]. This does > >> > not work anymore, we have to include ''/dev/'': disk = [ > >> > ''phy:/dev/vg/domU-root,sda1,w'' ].This seems like a bug in xl''s implementation of the syntax specification (docs/misc/xl-disk-configuration.txt), which states that any path which doesn''t start with a / should have "/dev/" prepended (IOW non-absolute paths are relative to /dev/). Anyone fancy knocking up a patch? Ian. _______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel
Hello, Am 11.04.2011 03:07, schrieb Sylvain CHEVALIER:> Many thanks Pasi and Claudiu for helping and sharing. > > I am sorry for not giving much more details about our experiences with > dom0 hanging, but I don''t want to touch our server right now as it is > running fine (except for qemu-dm memory leak) and people are using it. > I hope to set up another server this week and experiment with it. > These issues were similar to what Claudu described, but again, after > applying the fixes I described, it has been working fine, and we did > not have to revert it to 4.0. > > I think we had terminal issues, with any command run locally or > remotely in dom0 giving an ''input/output'' error message), and we could > not establish any other connection to it.I got here same Error after upgrading to xen-4.1 using 2.6.38-xen on gentoo. There are several domUs (gentoo, ubuntu, windows) running on this system it crashes without a kernel panic error just the ''input/output''-error. There are 4 NICs bonded to bond0 with 3 VLANs bond0.1, bond0.11, bond0.12. The bridges are xvbr1 >> bond0.1, xvbr11 >> bond0.11 and xvbr12 >> bond0.12. I DONT use the xen-bridge-script because there is no need to build the brigdes. RAM 48GB, 2x quadcore CPU (Intel Xeon). You said u fixed this issue with crashing, do u remember how? I am not really able to get informations out of all these "RE:" messages. I really would be happy if u could tell me u fixed it. Regards, Mike> > We are using the latest kernel on gentoo: 2.6.34-xen-r4 and had > similar issues with the previous one (2.6.34-xen-r3). > > Thanks, > > Sylvain > > On Mon, Apr 11, 2011 at 3:30 AM, Pasi Kärkkäinen<pasik@iki.fi> wrote: >> On Sun, Apr 10, 2011 at 12:56:05PM +0800, Sylvain CHEVALIER wrote: >>> Hi, >>> >>> We just migrated xen on gentoo on amd64 from 4.0 to 4.1 and >>> encountered a few issues that were not documented. Here is how we >>> solved them and what issues are remaining: >>> >>> 1. Our config files were referring to the LVM devices like >>> /dev/vg/domU-root with disk = [ ''phy:vg/domU-root,sda1,w'' ]. This does >>> not work anymore, we have to include ''/dev/'': disk = [ >>> ''phy:/dev/vg/domU-root,sda1,w'' ]. >>> >> >> Hello, >> >> I replied this 1. in another email. >> >>> 2. We were using bridges to map the 4 ethernet cards to virtual >>> machines so that virtual machines could have their own IP addresses. >>> In xen 4.1, it is recommended to configure bridges by ourselves and >>> not use xend although xend would still work. We could not get the >>> bridges to work and went on using xend. However, the network-bridge >>> script now includes a test: if a bridge already exist, then it would >>> do nothing. The problem with that is that our wrapper script for >>> network-bridge would only create the first bridge. So we had to >>> comment out the test in network-bridge, and it would work. Did anybody >>> successfully create bridges on gentoo? >>> Meanwhile, we found that if a domU is started and tries to access a >>> bridge that does not exist, dom0 would stop working. Shall it be a >>> bug? >>> >> >> What do you mean dom0 would stop working? Network dies? >> Does it work from the console? >> >> And yes, it''s recommended to configure networking using the distro >> network scripts, see here for examples: http://wiki.xen.org/xenwiki/HostConfiguration/Networking >> >> >>> 3. I believe we have memory issues when running a windows domU. We use >>> to allocate 512MB to domU and found that qemu would quickly use most >>> of it and after a while, dom0 would stop working. Now we allocate 4GB, >>> made sure there is no ballooning and it appears to work fine. Memory >>> leaks by qemu are widely reported but I wonder why we did not have >>> such issues before xen 4.1 update. >>> >> >> Again what do you mean with dom0 would stop working? dom0 kernel crash? hypervisor crash? >> And did you mean qemu-dm process in dom0 is leaking memory? >> >>> 4. We might also have issues with CPUs. dom0 and all domU were >>> authorized to use all CPUs, and we saw the whole system freezing often >>> and quickly when mounting in charge after the 4.1 update. After we >>> spinned one CPU to dom0 we haven''t seen this issue. Would this be a >>> bug? >>> >> >> This definitely sounds like a bug. >> >> Please set up a serial console and capture the error/crash messages: >> http://wiki.xensource.com/xenwiki/XenSerialConsole >> >> Also: what dom0 kernel version are you using? >> >>> Any hint or help on finding better solutions would be appreciated >>> >> >> -- Pasi >> >>_______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users