Hi there, I''ve been experimenting with iscsi and xen and have found an interesting behavior. I''d reasoned that using iscsi volumes as devices in a software RAID1 configuration would be a Good Idea since when one of the iscsi servers goes away the filesystem is stable. I have a dom0 which has access to two iscsi volumes on two seperate iscsi target servers. I configure a domU with these two iscsi targets as its block devices, ie: disk = [ ''phy:iscsi/iqn.2006-12.server1:raid/lun0/part,sda1,w'', ''phy:iscsi/iqn.2006-12.server2:raid/lun0/part,sdb1,w'' ] root = "/dev/md0 ro" This works in as much as the domU has root on a RAID1 md device and it functions normally; until I shutdown one of the iscsi servers. At that point the domU dies horribly. I see no messages on the domU console at all and so far nothing on the dom0 console either. Nothing in any of the logs. I also don''t see anything noteworthy in xm log nor xm dmesg when this happens; the domU becomes unresponsive and has to be killed with xm destroy. Am I expecting too much? Will xen just not cope when one of the block devices handed out to a domU simply ''goes away''? iscsi aside, what would xen be expected to do if you had two scsi drives, sda and sdb, presented to a domU as vbd''s and used by the domU in a RAID1 array... ie what would xen do if one of those hard drives failed? Thanks! _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
On 2/1/07, Steve Wray <steve.wray@cwa.co.nz> wrote:> Hi there, > I''ve been experimenting with iscsi and xen and have found an interesting > behavior. > > I''d reasoned that using iscsi volumes as devices in a software RAID1 > configuration would be a Good Idea since when one of the iscsi servers > goes away the filesystem is stable.Note todays thread "Linux software RAID1 (Dom0 or domU)?" .... jerry -- "Pay no attention to that man behind the curtain!" _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
Jerry Amundson wrote:> On 2/1/07, Steve Wray <steve.wray@cwa.co.nz> wrote: >> Hi there, >> I''ve been experimenting with iscsi and xen and have found an interesting >> behavior. >> >> I''d reasoned that using iscsi volumes as devices in a software RAID1 >> configuration would be a Good Idea since when one of the iscsi servers >> goes away the filesystem is stable. > > Note todays thread "Linux software RAID1 (Dom0 or domU)?" ....Ok I''ve been through that entire thread and it didn''t answer the question of what is happening to XEN when the device it presents to the domU as a vbd vanishes. This applies just as much to iscsi as to local disk; if I allocate a whole drive to a domU via XEN say something like: disk = [ ''phy:hdc,hda,w'', ''phy:hde,hdc,w'' ] what happens when hdc in dom0 dies horribly? Lets leave the issue of whats best practice WRT RAID for now and assume that the domU has hda and hdc as components of a RAID1 array so that, in principle, if the dom0 hdc fails everything *should* be ok; can XEN be expected to be ok with this? _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
Reasonably Related Threads
- Whether libvirt can support all backing chain layer are iscsi network disk type
- questions about disk auth element for Backing Chain
- Bug#691808: xcp-storage-managers: Another wrong binary path + wrong parameter in storage managers backend
- can't disconnec iSCSI targets, please help
- puppet/naginator?