Puppet in NFS fools Facter

[Forrie]

I have Puppet mounted to a couple of machines via NFS (along with
Ruby) -- and it works fine. Except, I just noticed that it fools
Facter in to believing it's a physical machine, when in fact one of
them is a VMware host.

We don't really use this setting, but I'm concerned other settings
might not be accurate.

Anyone know why this is happening?


Thanks.

[Russell Van Tassell]

Potentially related?

http://projects.puppetlabs.com/issues/10232

--
You received this message because you are subscribed to the Google Groups "Puppet Users" group.
To post to this group, send email to puppet...@googlegroups.com.
To unsubscribe from this group, send email to puppet-users...@googlegroups.com.
For more options, visit this group at http://groups.google.com/group/puppet-users?hl=en.

[Forrie]

I'm not sure I understand his setup, or what he means by "minimal
install".

My environment on the VMware image is CentOS 5.7, it is a full release
and the NFS mount contains a full release of Puppet and Ruby 1.8.x.

Perhaps there's something that Facter gets wrong when it's being
called from a non-system location?

Another possibility, is the previous run of Puppet was local -- each
machine still has a local /var/lib/puppet layout with all the
information that was stored (I'm not doing storedconfigs, yet).
Could information in there play into this somehow.


Thanks.

[Russell Van Tassell]

Any chance you copied over a bare metal box image, or something?

In any case, facter would seem to have only a handful of checks to determine "virtual" or non-virtual... unfortunately I don't have a vmware box in front of me to verify this, but you should be able to find facter's "virtual" tests in some place like:

/usr/lib/ruby/1.[89]/facter/util/virtual.rb

(slightly more convoluted if you're running puppet under rvm)

At first glance, for VMWare, it appears to be looking for /proc/self/status and/or /proc/virtual.

Thanks.

[Forrie]

Nothing was copied over. The NFS mount code was built and then
exported; 32- and 64-bit respectively. The code was built from
scratch and installed with the appropriate locally mounted prefix (in
this case, /local).

I'm on RHEL 5.x and we only have /proc/self/status which doesn't seem
to indicate anything about virtual or physical, for example:

Name: cat
State: R (running)
SleepAVG: 88%
Tgid: 1523
Pid: 1523
PPid: 31949
TracerPid: 0
Uid: 0 0 0 0
Gid: 0 0 0 0
FDSize: 256
Groups: 0 1 2 3 4 6 10
VmPeak: 58952 kB
VmSize: 58952 kB
VmLck: 0 kB
VmHWM: 484 kB
VmRSS: 484 kB
VmData: 176 kB
VmStk: 88 kB
VmExe: 20 kB
VmLib: 1448 kB
VmPTE: 40 kB
StaBrk: 06dd2000 kB
Brk: 06df3000 kB
StaStk: 7fffb84776a0 kB
Threads: 1
SigQ: 1/4096
SigPnd: 0000000000000000
ShdPnd: 0000000000000000
SigBlk: 0000000000000000
SigIgn: 0000000000000000
SigCgt: 0000000000000000
CapInh: 0000000000000000
CapPrm: 00000000fffffeff
CapEff: 00000000fffffeff
Cpus_allowed:
00000000,00000000,00000000,00000000,00000000,00000000,00000000,00000003
Mems_allowed: 00000000,00000001

[Forrie]

Interestingly, the command "facter serialnumber" correctly pulls that
it's a VMware system:

# /local/bin/facter serialnumber
VMware-56 4d 00 7e e8 3b e8 c9-85 7f 4e XX XX XX XX XX

[Forrie]

On another system, same NFS mounts, the "facter virtual" reports the
correct information, that system is running:

2.6.18-194.3.1.el5

The system that doesn't correctly report is:

2.6.18-274.18.1.el5

I don't know if that really matters.