If you want to do stateful it's much handier to go the golden image route. Build a node up the way you want it, then grab the image using rsync, create vnfs/bootstrap, define the node object in warewulf and off you go. My progress is detailed at
https://dokuwiki.wesleyan.edu/doku.php?id=cluster:144
-Henk
I was following the Centos-Warewulf-PBSPro recipe (https://github.com/openhpc/ohpc/releases/download/v1.3.2.GA/Install_guide-CentOS7-Warewulf-PBSPro-1.3.2-x86_64.pdf) from OpenHPC. My issue turned out to be several typos introduced while setting up the node in warewulf. It is working now and I have duplicated the steps for a statefull client node as well.
Cheers.
--JohnM
From: itoufique [mailto:tech...@gmail.com]
Sent: Wednesday, 13 September 2017 15:35
To: Warewulf <ware...@lbl.gov>
Subject: [Warewulf] Re: Stateful provisioning not working
Hi Guys,
--
[root@b0 ~]# fdisk -l Disk /dev/sda: 80.0 GB, 80026361856 bytes 255 heads, 63 sectors/track, 9729 cylinders Units = cylinders of 16065 * 512 = 8225280 bytes Sector size (logical/physical): 512 bytes / 512 bytes I/O size (minimum/optimal): 512 bytes / 512 bytes Disk identifier: 0xcece6336 Device Boot Start End Blocks Id System /dev/sda1 * 1 64 512000 83 Linux Partition 1 does not end on cylinder boundary. /dev/sda2 64 319 2048000 82 Linux swap / Solaris Partition 2 does not end on cylinder boundary. /dev/sda3
#1 I'd try to keep my OS disks tiny and use the large disks for /scratch, helps with reimaging speeds
#2 Are you sure it's not there? Commands like df/mount will not show it, try 'free -m' on the node
You may also want to try swapon /dev/sda2, if that works stick it in /etc/rc.local
-Henk
Hi John,
I used the same recipe to build the cluster. Unfortunately, i am getting the same error. Please can you post what typos did you find?
Thanks
To view this discussion on the web visit https://groups.google.com/a/lbl.gov/d/msgid/warewulf/2346db6f-38d7-4b7f-a699-e165a8cc68f9%40lbl.gov.
For more options, visit https://groups.google.com/a/lbl.gov/d/optout.