Hello, I''ve been testing live migrations with Xen 4.1 and everything seems to be going as it should except after the migration succeeds, the domU loses access to the network. I''m using ubuntu 11.10 as the dom0 and as the domU, kernels were all gotten through apt (3.0.0-12). I know the migration succeeds because I can still ''xm console'' into the domU and I can see processes I started before the migration are still running. If I reboot the vm, the networking situation corrects itself and the domU regains access to the network, but then it''s not really ''live'' which is important to me. There are two things that stick out at me after looking at all the logs, dmesg etc.: 1. The vif after migration seems to have the ''tx-/rx-ring-ref'' stuck reversed from what it was originally on the pre-migration dom0. After a domU reboot, the tx-/rx-ring-ref goes back to the original ordering, for example, an ''xm network-list'' on the domU shows this just after the migration: Idx BE MAC Addr. handle state evt-ch tx-/rx-ring-ref BE-path 0 0 00:16:3e:17:ed:97 0 4 10 769 /768 /local/domain/0/backend/vif/18/0 And then after a reboot, I regain access and this is what it looks like: Idx BE MAC Addr. handle state evt-ch tx-/rx-ring-ref BE-path 0 0 00:16:3e:17:ed:97 0 4 10 768 /769 /local/domain/0/backend/vif/18/0 2. The only error message I can pick out from the xend.log is this on the original pre-migration dom0: [2012-02-07 20:53:20 1278] DEBUG (XendCheckpoint:124) [xc_save]: /usr/lib/xen-4.1/bin/xc_save 20 126 0 0 1 [2012-02-07 20:53:20 1278] INFO (XendCheckpoint:423) xc_save: failed to get the suspend evtchn port However, everything migrates successfully other than this network issue. I can''t really figure out the significance of either issue after a lot of googling. Any insight would be greatly appreciated if anyone has run into anything like this or has a hunch. Thanks, Chris _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
Also after searching it seems that Dave Nuzzo had exactly the same problem with the same setup: http://old-list-archives.xen.org/archives/html/xen-users/2011-10/msg00491.html I just joined the list and his email address is hidden on that public page. Is there some way I can contact him directly to see if he solved it? Thanks, Chris On Tue, Feb 7, 2012 at 4:25 PM, Chris Dickson <chrisd1100@gmail.com> wrote:> Hello, > > I''ve been testing live migrations with Xen 4.1 and everything seems to be > going as it should except after the migration succeeds, the domU loses > access to the network. I''m using ubuntu 11.10 as the dom0 and as the domU, > kernels were all gotten through apt (3.0.0-12). I know the migration > succeeds because I can still ''xm console'' into the domU and I can see > processes I started before the migration are still running. If I reboot the > vm, the networking situation corrects itself and the domU regains access to > the network, but then it''s not really ''live'' which is important to me. > There are two things that stick out at me after looking at all the logs, > dmesg etc.: > > 1. The vif after migration seems to have the ''tx-/rx-ring-ref'' stuck > reversed from what it was originally on the pre-migration dom0. After a > domU reboot, the tx-/rx-ring-ref goes back to the original ordering, for > example, an ''xm network-list'' on the domU shows this just after the > migration: > > Idx BE MAC Addr. handle state evt-ch tx-/rx-ring-ref BE-path > 0 0 00:16:3e:17:ed:97 0 4 10 769 /768 > /local/domain/0/backend/vif/18/0 > > And then after a reboot, I regain access and this is what it looks like: > > Idx BE MAC Addr. handle state evt-ch tx-/rx-ring-ref BE-path > 0 0 00:16:3e:17:ed:97 0 4 10 768 /769 > /local/domain/0/backend/vif/18/0 > > 2. The only error message I can pick out from the xend.log is this on the > original pre-migration dom0: > > [2012-02-07 20:53:20 1278] DEBUG (XendCheckpoint:124) [xc_save]: > /usr/lib/xen-4.1/bin/xc_save 20 126 0 0 1 > [2012-02-07 20:53:20 1278] INFO (XendCheckpoint:423) xc_save: failed to > get the suspend evtchn port > > However, everything migrates successfully other than this network issue. > > I can''t really figure out the significance of either issue after a lot of > googling. Any insight would be greatly appreciated if anyone has run into > anything like this or has a hunch. > > Thanks, > > Chris >_______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
Update: I installed the ubuntu kernel package ''linux-image-extra-3.0.0-15-virtual'' on all the domUs and dom0s in question and the issue is gone. Who knows exactly what it was. Chris On Tue, Feb 7, 2012 at 7:02 PM, Chris Dickson <chrisd1100@gmail.com> wrote:> Also after searching it seems that Dave Nuzzo had exactly the same problem > with the same setup: > > > http://old-list-archives.xen.org/archives/html/xen-users/2011-10/msg00491.html > > > I just joined the list and his email address is hidden on that public > page. Is there some way I can contact him directly to see if he solved it? > > Thanks, > > Chris > > > On Tue, Feb 7, 2012 at 4:25 PM, Chris Dickson <chrisd1100@gmail.com>wrote: > >> Hello, >> >> I''ve been testing live migrations with Xen 4.1 and everything seems to be >> going as it should except after the migration succeeds, the domU loses >> access to the network. I''m using ubuntu 11.10 as the dom0 and as the domU, >> kernels were all gotten through apt (3.0.0-12). I know the migration >> succeeds because I can still ''xm console'' into the domU and I can see >> processes I started before the migration are still running. If I reboot the >> vm, the networking situation corrects itself and the domU regains access to >> the network, but then it''s not really ''live'' which is important to me. >> There are two things that stick out at me after looking at all the logs, >> dmesg etc.: >> >> 1. The vif after migration seems to have the ''tx-/rx-ring-ref'' stuck >> reversed from what it was originally on the pre-migration dom0. After a >> domU reboot, the tx-/rx-ring-ref goes back to the original ordering, for >> example, an ''xm network-list'' on the domU shows this just after the >> migration: >> >> Idx BE MAC Addr. handle state evt-ch tx-/rx-ring-ref BE-path >> 0 0 00:16:3e:17:ed:97 0 4 10 769 /768 >> /local/domain/0/backend/vif/18/0 >> >> And then after a reboot, I regain access and this is what it looks like: >> >> Idx BE MAC Addr. handle state evt-ch tx-/rx-ring-ref BE-path >> 0 0 00:16:3e:17:ed:97 0 4 10 768 /769 >> /local/domain/0/backend/vif/18/0 >> >> 2. The only error message I can pick out from the xend.log is this on the >> original pre-migration dom0: >> >> [2012-02-07 20:53:20 1278] DEBUG (XendCheckpoint:124) [xc_save]: >> /usr/lib/xen-4.1/bin/xc_save 20 126 0 0 1 >> [2012-02-07 20:53:20 1278] INFO (XendCheckpoint:423) xc_save: failed to >> get the suspend evtchn port >> >> However, everything migrates successfully other than this network issue. >> >> I can''t really figure out the significance of either issue after a lot of >> googling. Any insight would be greatly appreciated if anyone has run into >> anything like this or has a hunch. >> >> Thanks, >> >> Chris >> > >_______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users