Hi all! I''m using AoE to export a ReiserFS partition that contains a file-backed VBDs of a DomU to perform tests of migration between two xen host. The three systems have Debian GNU/Linux Etch. Both writes and deletes of files in the file-backed VBD DomU works synchronized, but the problem seems to be when I want to create a file in the exported partition together with the file-backed VBDs DomU. In this case, doesn''t seem syncronize with the rest of the both import and export servers. Is it possible to perform this writes and deletes in synchronized mode too? Thanks in advance. Regards, Daniel -- Daniel Bareiro - System Administrator Fingerprint: BFB3 08D6 B4D1 31B2 72B9 29CE 6696 BF1B 14E6 1D37 Powered by Debian GNU/Linux Etch - Linux user #188.598 _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
On Saturday 02 December 2006 4:06 pm, Daniel Bareiro wrote:> I''m using AoE to export a ReiserFS partition that contains a file-backed > VBDs of a DomU to perform tests of migration between two xen host. The > three systems have Debian GNU/Linux Etch.are you mounting the same volume on two hosts? remember that ReiserFS isn''t a cluster filesystem, and it won''t stay consistent. IOW, this is a sure recipe to corrupt the FS -- Javier _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
On Saturday, 02 December 2006 19:18:40 -0500, Javier Guerra wrote: Hi Javier. Thanks for your answer.> > I''m using AoE to export a ReiserFS partition that contains a > > file-backed VBDs of a DomU to perform tests of migration between two > > xen host. The three systems have Debian GNU/Linux Etch.> are you mounting the same volume on two hosts? remember that ReiserFS > isn''t a cluster filesystem, and it won''t stay consistent. IOW, this is > a sure recipe to corrupt the FSYes. I must mount /dev/etherd/e0.0 on two host to perform the migration, because of the file-backed VBD of DomU resides in this exported partition. How I said, both files writes and deletes in the file-backed VBD of DomU works synchronized, but the problem seems to be when I want to create a file in the exported partition together with the file-backed VBDs of DomU. Why this difference in the synchronization? How I get writes and deletes in synchronized mode to both file-backed VBD of DomU and others files in the same exported partition? Regards, Daniel -- Daniel Bareiro - System Administrator Fingerprint: BFB3 08D6 B4D1 31B2 72B9 29CE 6696 BF1B 14E6 1D37 Powered by Debian GNU/Linux Etch - Linux user #188.598 _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
On Dec 2, 2006, at 6:30 PM, Daniel Bareiro wrote:> On Saturday, 02 December 2006 19:18:40 -0500, > > Hi Javier. Thanks for your answer. > >>> I''m using AoE to export a ReiserFS partition that contains a >>> file-backed VBDs of a DomU to perform tests of migration between two >>> xen host. The three systems have Debian GNU/Linux Etch. > >> are you mounting the same volume on two hosts? remember that >> ReiserFS >> isn''t a cluster filesystem, and it won''t stay consistent. IOW, >> this is >> a sure recipe to corrupt the FS > > Yes. I must mount /dev/etherd/e0.0 on two host to perform the > migration, > because of the file-backed VBD of DomU resides in this exported > partition. How I said, both files writes and deletes in the file- > backed > VBD of DomU works synchronized, but the problem seems to be when I > want > to create a file in the exported partition together with the file- > backed > VBDs of DomU. Why this difference in the synchronization? How I get > writes and deletes in synchronized mode to both file-backed VBD of > DomU > and others files in the same exported partition?What Javier said is correct. What you are trying to do *will not work*. I cannot answer your question specifically, but I can say without question that the weirdness you''re seeing is a side-effect of the root problem, and that you''ll have bigger problems than you''re already seen, as in data loss and kernel panics, if you continue down this road. -- -- Tom Mornini, CTO -- Engine Yard, Ruby on Rails Hosting -- Reliability, Ease of Use, Scalability -- (866) 518-YARD (9273) _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
On Sat, 2006-12-02 at 18:06 -0300, Daniel Bareiro wrote:> Hi all! > > I''m using AoE to export a ReiserFS partition that contains a file-backed > VBDs of a DomU to perform tests of migration between two xen host. The > three systems have Debian GNU/Linux Etch.You''d do better off using LVM backed VBD''s and exporting them individually. Unless your (main) holding partition is using a cluster FS, this will lead to a melt down. You would need to use something like ocfs2 (easy) or gfs (a little trickier) or clvm (inbetween) to accomplish this. The partition that holds the file backed VBD''s *must* be a cluster file system.> > Both writes and deletes of files in the file-backed VBD DomU works > synchronized, but the problem seems to be when I want to create a file > in the exported partition together with the file-backed VBDs DomU. In > this case, doesn''t seem syncronize with the rest of the both import and > export servers. >Again, you can not mount (in rw mode) *any* partition that is not formatted as a cluster file system from two computers.> Is it possible to perform this writes and deletes in synchronized mode too? > > Thanks in advance. >I''d just make smaller volume groups, and copy over the loops to their respective new homes (the smaller VG''s).> Regards, > Daniel > _______________________________________________ > 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
On Saturday, 02 December 2006 19:07:45 -0800, Tom Mornini wrote:> >>are you mounting the same volume on two hosts? remember that > >>ReiserFS isn''t a cluster filesystem, and it won''t stay consistent. > >>IOW, this is a sure recipe to corrupt the FS> >Yes. I must mount /dev/etherd/e0.0 on two host to perform the > >migration, because of the file-backed VBD of DomU resides in this > >exported partition. How I said, both files writes and deletes in the > >file-backed VBD of DomU works synchronized, but the problem seems to > >be when I want to create a file in the exported partition together > >with the file-backed VBDs of DomU. Why this difference in the > >synchronization? How I get writes and deletes in synchronized mode to > >both file-backed VBD of DomU and others files in the same exported > >partition?> What Javier said is correct. What you are trying to do *will not > work*.> I cannot answer your question specifically, but I can say without > question that the weirdness you''re seeing is a side-effect of the root > problem, and that you''ll have bigger problems than you''re already > seen, as in data loss and kernel panics, if you continue down this > road.Mmmm... I see. Thanks for the warning. For the moment, I''ll analyze what Tim said. But all recommendation are welcome. Regards, Daniel -- Daniel Bareiro - System Administrator Fingerprint: BFB3 08D6 B4D1 31B2 72B9 29CE 6696 BF1B 14E6 1D37 Powered by Debian GNU/Linux Etch - Linux user #188.598 _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
On Sunday, 03 December 2006 13:24:21 +0800, Tim Post wrote:> > I''m using AoE to export a ReiserFS partition that contains a > > file-backed VBDs of a DomU to perform tests of migration between two > > xen host. The three systems have Debian GNU/Linux Etch.> You''d do better off using LVM backed VBD''s and exporting them > individually.What do you refer with *individually*? I thought to export a partition (PV) using vblade. This partition contains the LV of the domU''s FS. Then, in the Xen hosts, I import the device using aoe module/vgscan/vgchange -a y. Now then, as long as the simultaneous mounting ---to perform the migration--- in the two Xen host can corrupt the FS, must I put the device in the domU config file? Something like: disk = [ ''phy:vg_xen/lv_vm1,hda1,w'' ] Maybe, another choice is to use losetup to associate a loop device with the file-backed VBD of DomU. Then, we export the loop device. I guess that the VBD configuration lines would change to: disk = [ ''phy:etherd/e0.0,hda1,w'' ] Is it correct?> > Both writes and deletes of files in the file-backed VBD DomU works > > synchronized, but the problem seems to be when I want to create a > > file in the exported partition together with the file-backed VBDs > > DomU. In this case, doesn''t seem syncronize with the rest of the > > both import and export servers.> Again, you can not mount (in rw mode) *any* partition that is not > formatted as a cluster file system from two computers.Do you know if the kernel provided for Xensources supports the cluster file systems that you''ve mentioned? This file systems must be used together with AoE as well? What about csync2 [1]?> > Is it possible to perform this writes and deletes in synchronized > > mode too?> I''d just make smaller volume groups, and copy over the loops to their > respective new homes (the smaller VG''s).Do you refer to copy the content of the file-backed VBD to a LV into the VG? Thanks for your answer. Regards, Daniel [1] http://oss.linbit.com/csync2/ -- Daniel Bareiro - System Administrator Fingerprint: BFB3 08D6 B4D1 31B2 72B9 29CE 6696 BF1B 14E6 1D37 Powered by Debian GNU/Linux Etch - Linux user #188.598 _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users