Re: Xen 4.2 added to pkgsrc

[jnemeth%victoria.tc.ca@localhost (John Nemeth)]

On Oct 15,  7:46pm, David Brownlee wrote:
} On 25 May 2013 06:33, John Nemeth <jnemeth%victoria.tc.ca@localhost> wrote:
} > On Oct 11,  4:03am, David Brownlee wrote:
} > }  On 15 May 2013 08:18, John Nemeth <jnemeth%victoria.tc.ca@localhost> 
wrote:
} > } >      I've just committed sysutils/xenkernel42 and sysutils/xentools42.
} > } > They have had fairly extensive testing, but they are just initial
} > } > commits.  I've added them so that people can test them, break them,
} > } > improve them, etc.
} > }
} > } [some test feedback & 'fixed' replies from John]
} >
} > } /var/run/xen/bootloader.1.d was not present - created by hand to continue
} >
} >      Not sure what this is, or why it was needed?
} 
} Unknown. All I know is that I needed to create it to get xen to work.
} If I recall /var/run/xen was also missing,

     I added /var/run/xen to the list of directories to create.

} > } On starting the guests (CentOS 6) they seemed to fire up ok but then 
exited.
} > }
} > } The console had:
} > } XENBUS: Waiting for devices to initialise:
} > } [...]
} > } XENBUS: Timeout connecting to device: device/vbd/51712 (local state 3,
} > } remote state 2)
} >
} > } the .cfg for that VM:
} > } name = "jira"
} > } memory = "1536"
} > } disk = [ 'file:/opt/xen/img/jira.img,0xca00,w', ]
} > } vif = [ 'mac=3D02:00:00:00:01:03, bridge=bridge0', ]
} > } bootloader = "/usr/pkg/bin/pygrub"
} > } vcpus = 3
} >
} >      Why are you using a hexadecimal number for your disk device?  For
} > an PV linux domU, my disks look like:
} 
} "Laughing dried grapes" [*]. Quite possibly I found it was needed in

     I don't see the "[*]", but after thinking about it, I got it.

} xen40, and its not been changed since. Am building xen42 again and
} will retest with conventional names.

     I don't recall ever using hexadecimal numbers for disk names, and
I've played with a number of different versions of Xen, but whatever...

} > #disk = [ 'phy:/dev/mapper/vg0-domu--fedora,xvda,w' ]
} > #disk = [ 'phy:/dev/mapper/vg0-domu--fedora,xvda,w', 
'file:/usr/local/isos/Fedora-17-x86_64-DVD.iso,xvdb,r' ]
} > disk = [ 'phy:/dev/mapper/vg0-domu--fedora,xvda,w', 
'file:/usr/local/isos/Fedora-18-Beta-x86_64-DVD.iso,xvdb,r' ]
} > #disk = [ 'file:/mnt/fedora,xvda,w' ]
} >
} > And, for an HVM linux domU, my disks look like:
} >
} > disk = [ 'phy:/dev/mapper/vg0-domu--hvmlinux,ioemu:hda,w' ]
} > #disk = [ 'phy:/dev/mapper/vg0-domu--hvmlinux,ioemu:hda,w',
} > #         'file:/usr/local/isos/hvmlinux-6.0rc12.iso,ioemu:hdc:cdrom,r' ]
} >
} > Notice that I use xvda and xvdb, or hda and hdc, depending on the type
} > of client.  Also, do you have enough vnds (i.e. does vnconfig -l show
} > any empty slots).  When I setup a Xen system with a bunch of domUs
} > using file based disks, I found that I had manually MAKEDEV a bunch of
} > vnds.  The problem symptom was that the domU couldn't get a disk.  I'm
} > assuming that jira.img isn't in use by anything else.
} 
} Exactly the same config works fine under xen41 (I have 10 vnd devices
} as hit that limit there)

     Perhaps, but it would still be good to make sure there are some
free.  The only things I ses that are interesting are the weird disk
names and the possibility of running out of vnds.  Otherwise, I don't
know what the issue could be and it would need a lot more
investigating.

} > } Rolling back to 41 for now :)
} >
} >      That's too bad.  Maybe this update will smooth out some of the issues.
} 
} Thanks for the update - rebuilding xen42 packages now and will retest
} in the morning :)

     I await the new report with baited breathe.  :-)

}-- End of excerpt from David Brownlee