#4654 mixed cluster, create stateless/diskless image over NFS mounted /install fails

[Victor Hu]

Under the mixed cluster support, there's a section on building stateless images for different architecture or OS.

https://sourceforge.net/p/xcat/wiki/Building_a_Stateless_Image_of_a_Different_Architecture_or_OS/

I am trying to use a RHELS 6.6 Mgmt node and be able to deploy diskless images for other OS (Target: RHELS 7.1). On a RHELS 7.1 compute node, I followed the instructions and mounted the /install filesystem and scp over the files under /opt/xcat/share/xcat/netboot

Using the command generated from the genimgae --dryrun, I execute that on the RHELS 7.1 node. It's slow since it's writing to NFS mounted /install directory.

While installing, I hit the following errors:

...
Installing : systemd-208-20.el7.ppc64 514/824
Error unpacking rpm package systemd-208-20.el7.ppc64
error: unpacking of archive failed on file /usr/bin/systemd-detect-virt: cpio: cap_set_file
error: systemd-208-20.el7.ppc64: install failed
...
...
Installing : iputils-20121221-6.el7.ppc64 543/824
Error unpacking rpm package iputils-20121221-6.el7.ppc64
error: unpacking of archive failed on file /usr/bin/ping: cpio: cap_set_file
error: iputils-20121221-6.el7.ppc64: install failed
Installing : initscripts-9.49.24-1.el7.ppc64 544/824
Error unpacking rpm package initscripts-9.49.24-1.el7.ppc64
error: unpacking of archive failed on file /usr/lib/systemd/system/basic.target.wants/rhel-autorelabel-mark.service;55355c91: cpio: symlink
error: initscripts-9.49.24-1.el7.ppc64: install failed
...

so the genimage does not finish successfully...

I did put the location to /root (on the native filesystem) and the genimage ran with no issues. So Maybe there's something that we need to do for the NFS mount options? Not sure.

[Guang Cheng Li]

We got this issue reported from a customer the last week, he tried to setup PCM HA environment using NFS mounted /install, after some investigation, it turned out to be a NFS/rpm problem, I opened the RedHat bug earlier this week, has not got useful hints from RedHat yet. For the mixed cluster, if your MN/CN is the same Linux distribution and the same arch, the os version does not matter, you should be able to generate RHEL6 diskless image on RHEL 7 MN or generate RHEL 7 diskless image on RHEL 6 MN.

[Victor Hu]

Thanks Guang Cheng, It looks like I was able to get around this by pointing the rootimgdir to a local directory , then tar up the directory and scp back to the mgmt node so I'm wondering if this is a case of doing chroot in a mounted NFS environment?

I see many of those messages appear doing a google search but it's related to docker containers and other various container environments.

[Guang Cheng Li]

Victor, it is right that the xCAT diskless support has quite a bit similarities with docker, they actually uses the same concept of chroot environment, I also noticed some common issues between xCAT diskless and docker.

For getting around this issue, what I used was imgexport and imgimport, create the image on another machine, then imgexport, copy the bundle .tgz file under /root/ to the MN, then use imgimport to import the image. The imgimport/imgexport will take care of the osimage configuration files like kickstart template, pkglist and otherpkglist,etc besides the rootimg itself.

[Guang Cheng Li]

The bugzilla number is 124177, using this bug to track the RedHat bugzilla bug