Thanks, Chris! Yeah, that's a good point about reinstall. I had read the docs, but for some reason reset wasn't working the way I thought it should, but I've since found that it was a misunderstanding on my part. I've put it back to "reset" and I think that's giving me more of what I want.
Thanks for the pointer to control.sh. dd'ing /dev/sda and /dev/sda5 (which happens to be our logical volume disk in this case) did the trick for me, but I'll probably do something more similar to what you're doing in control.sh; except that I used -bs=1M and -count=100 (that was enough for me to convince the node to PXE boot and wipe clean what needed to be clean). I very much appreciate your response and help! Best, Jeff From: [email protected] [mailto:[email protected]] Sent: Tuesday, April 23, 2013 12:56 PM To: Walls, Jeffrey Joel (HP Converged Cloud - Cloud OS); [email protected] Subject: RE: "Reset Node" - Volume group name already in use Hey Jeff, If you want to repurpose a node that has been allocated and had some proposals applied to it, then you should remove the node from any policies that it is in, and then delete the node. When you reboot it, it should PXE boot and be discovered again. At that point, you should be able to allocate it and apply different proposals to it. We have seen some cases where if you have a hardware RAID set up, that the system won't PXE boot. In this case, you can go into the RAID controller firmware during the POST, nuke the RAID config, and then reboot to get it to PXE boot. We have encountered rare times when this doesn't work, and in those cases have to resort to dd'ing a chunk of the disk to zeros (including the boot sector) just as you did. This is slow though, so we only do this as a last resort. See the nuke_everything method in crowbar/barclamps/provisioner/updates/control.sh. The reinstall command is not what you want in this case. From the User's Guide: "Reinstall will reinstall the base OS on the node, as well as redeploying the roles applied to the node.". So basically a reinstall will slick the node, then reinstall the OS and all the software that was installed on it as a result of including the node in whatever proposals it's in. Thanks! Chris Crowbar Developer Dell From: crowbar-bounces On Behalf Of Walls, Jeffrey Joel (HP Converged Cloud - Cloud OS) Sent: Friday, April 19, 2013 9:12 PM To: crowbar Subject: Re: [Crowbar] "Reset Node" - Volume group name already in use I think I've figured it out. It may be completely wrong, but the way I was causing the PXE boot was to zero the boot sector of the node before the reboot. I increased that to 10Mb to make sure the partition table was zeroed as well, and now the process seems to work. I'm still interested if anyone else has encountered something like this and what a better solution might be. Thanks! Jeff From: [email protected]<mailto:[email protected]> [mailto:[email protected]] On Behalf Of Walls, Jeffrey Joel (HP Converged Cloud - Cloud OS) Sent: Friday, April 19, 2013 3:27 PM To: [email protected]<mailto:[email protected]> Subject: [Crowbar] "Reset Node" - Volume group name already in use Hi Crowbar'ers, I apologize in advance if this is a dumb question, but I'm stumped. I've dug through logs, read source, googled and binged and can't seem to find an answer that moves me forward. If you can help, I'd greatly appreciate it. I have a pool of VM's running in a VMWare vSphere cluster. Crowbar is running fine and PXE booted these nodes and was able to allocate them. Note that the "bios" and "raid" are both set to "not set" on all of these nodes. I then applied some barclamps and that worked fine as well. I want to return some of these nodes to an unallocated state (with all of the "junk" that the barclamps put down removed), so I issued the command: crowbar_machines reinstall <node_id> And then forced the node to PXE-boot. We're not using BMC, so I had to force the node to PXE-boot, but I presume that's ok? It PXE-boots fine and then enters the loop waiting to be allocated. When I later allocate the node, it fails booting Ubuntu with the message: Partition disks - Volume group name already in use. I've tried "reset", "reinstall", and "delete" and none of them give me the behavior I want, which is just to "give me a fresh new system." So my questions are: 1) Do I have the process correct for "returning a node to an unallocated state"? 2) Why would the volume group still be set after a PXE boot? 3) What else can I do to track this down? 4) What additional information can I provide to help? Thanks for any help you can give! Jeffrey Walls HP Converged Cloud Cloud OS [email protected]<mailto:[email protected]> T +1 970 898 1619 Hewlett-Packard Company 3404 E. Harmony Road, MS-F Fort Collins, CO 80525 USA
_______________________________________________ Crowbar mailing list [email protected] https://lists.us.dell.com/mailman/listinfo/crowbar For more information: http://crowbar.github.com/
