Hi. My SuSE 10.3 system''s hard disk died, so I opted for a bigger hard disk as a replacement for xmas. My xen system is Fedora fc8, since that system supports hvm. My SuSE system is formatted with lvm, and comes with nfs. 1) Is nfs sufficient for serving up file-backed stores, or is the performance difference with ocfs2 or gfs that much greater? 2) If nfs is sufficient, version 3 or 4? In other words, does xen do record locking (only available in 4), or just file locking? My previous non-xen experience is that sqlite databases only work on nfs4. 3) I don''t suppose I can export a logical volume device to my xen server, right? So I''m stuck with creating file backed stores, right? Thanx. _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
jim burns wrote:> Hi. My SuSE 10.3 system''s hard disk died, so I opted for a bigger hard disk as > a replacement for xmas. My xen system is Fedora fc8, since that system > supports hvm. My SuSE system is formatted with lvm, and comes with nfs. > > 1) Is nfs sufficient for serving up file-backed stores, or is the performance > difference with ocfs2 or gfs that much greater? >NFS is "stateless". This is begging, begging, begging for pain if used as a file-based Xen image storage. Can you go to iscsi?> 2) If nfs is sufficient, version 3 or 4? In other words, does xen do record > locking (only available in 4), or just file locking? My previous non-xen > experience is that sqlite databases only work on nfs4. > > 3) I don''t suppose I can export a logical volume device to my xen server, > right? So I''m stuck with creating file backed stores, right? >Whyever can''t you export logical volume to your Xen server? Do you mean an LVM type volume? I certainly use those! _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
jim burns wrote:> Hi. My SuSE 10.3 system''s hard disk died, so I opted for a bigger hard > disk as a replacement for xmas. My xen system is Fedora fc8, since that > system supports hvm. My SuSE system is formatted with lvm, and comes with > nfs. > > 1) Is nfs sufficient for serving up file-backed stores, or is the > performance difference with ocfs2 or gfs that much greater?I dont like file backed stores that much, same goes for nfs. nfs and big files gave me lots of trouble in the past.> > 2) If nfs is sufficient, version 3 or 4? In other words, does xen do > record locking (only available in 4), or just file locking? My previous > non-xen experience is that sqlite databases only work on nfs4.> > 3) I don''t suppose I can export a logical volume device to my xen server, > right? So I''m stuck with creating file backed stores, right?Have a look at gnbd, iscsi or ataoe, i use iscsi and it works like a charm.> > Thanx.Welcome ;) _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
On Mon December 24 2007 12:07:03 pm Nico Kadel-Garcia wrote:> NFS is "stateless". This is begging, begging, begging for pain if used > as a file-based Xen image storage. Can you go to iscsi?Thanx for your response. Could you be more explicit why statefull is better? Maybe some war stories :-)> Whyever can''t you export logical volume to your Xen server? Do you mean > an LVM type volume? I certainly use those!Since you apparently don''t use nfs, I''m not sure how you do that. Could you post the relevent config files? Thanx. _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
On Mon December 24 2007 5:54:41 pm Bastiaan wrote:> Have a look at gnbd, iscsi or ataoe, i use iscsi and it works like a charm.Thanx. Do you have good urls for these options? I''m looking for software solutions only. New hardware is not currently in the budget. _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
On Tue December 25 2007 4:44:43 am you wrote:> It''s not clear what you mean by "export" of volumes. I''ve been working > extensively with local LVM partitions on the server itself, not network > exporting them. I''ve been longing to but lacked the opportunity to do > iscsi based mounting to permit live migrations.''network exporting'' is precisely what I would like to do, since I have very little more space on my xen server for anything more than proof of concept vms, and I have so much space now on my SuSE system. My experience with nfs 3 or 4 is you can only export previously mounted portions of the disk, which means effectively you have nfs -> file backed store -> /dev/mapper/imagefile, and I was hoping to remove some of the overhead. By the way, I found the following urls on the OpenSuSE wiki, but nothing on gnbd or ataoe. I suppose I can google for the their home pages, but I''m not looking to pioneer technology on SuSE that hasn''t been trod elsewhere :-; http://en.opensuse.org/Open-iSCSI_and_SUSE_Linux http://www.open-iscsi.org/ _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
On Tuesday 25 December 2007, jim burns wrote:> ''network exporting'' is precisely what I would like to do, since I have very > little more space on my xen server for anything more than proof of concept > vms, and I have so much space now on my SuSE system. My experience with nfs > 3 or 4 is you can only export previously mounted portions of the disk, > which means effectively you have nfs -> file backed store -> > /dev/mapper/imagefile, and I was hoping to remove some of the overhead.NFS is a filesystem server protocol; therefore you can use it to host a file backed store. to use blockdevice backed stores you need a blockdevice sharing protocol. the most popular are iSCSI, AoE and gnbd other kind of solution is using a shared block device (via iSCSI, FC, IEEE1394, AoE, gnbd, drbd, etc) and put a cluster filesystem on it (GFS, OCFS2, etc) and put image files there. it would give you the easy of filebacking without the overhead of filesharing (far higher than block sharing). i haven''t used it, and don''t think that filebacking is so much easier to handle than blockdevice stores; but some people here like the setup. -- Javier _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
On Tue December 25 2007 9:20:44 pm Javier Guerra Giraldez wrote:> NFS is a filesystem server protocol; therefore you can use it to host a > file backed store. to use blockdevice backed stores you need a blockdevice > sharing protocol. the most popular are iSCSI, AoE and gnbdOk - that clears it up. Any favorite urls? Thanx. _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
jim burns wrote:> On Tue December 25 2007 9:20:44 pm Javier Guerra Giraldez wrote: >> NFS is a filesystem server protocol; therefore you can use it to host a >> file backed store. to use blockdevice backed stores you need a >> blockdevice sharing protocol. the most popular are iSCSI, AoE and gnbd > > Ok - that clears it up. Any favorite urls? Thanx.http://lists.linbit.com/pipermail/drbd-user/2007-January/006290.html http://www.gridvm.org/drbd-lvm-gnbd-and-xen-for-free-and-reliable-san.html http://lists.xensource.com/archives/html/xen-devel/2005-06/msg00556.html http://wiki.novell.com/images/0/01/CHASF_preview_Nov172006.pdf https://open.datacore.ch/DCwiki.open/Wiki.jsp?page=GFS.Install#section-GFS.Install-InstallationOfGFSOnASystemRunningTheGenTooLinuxDistributionhttpgentoo.org These are the pages i got most info from. What i am currently trying are: 1. 2 boxes, with a master-master drbd block device replication, 2. evms cluster aware volume manager 3. iSCSI targets on both hosts 4. multipath iscsi setup on the xen servers (seperate boxes). one and two are working, three does work, but needs ''double configuration'' (configuration on both hosts), which i don''t like. _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users