Menu

#100 Wrong "Guest OS" information in vSphere client displayed

closed-wont-fix
nobody
None
5
2011-10-31
2011-05-05
Timo Gurr
No

The guest is configured as "Other 2.6x Linux (64-bit)", attached screenshot (vmware-tools_stopped.png).

# uname -a
Linux wwwproxy 2.6.38.5 #2 SMP Wed May 4 12:33:49 CEST 2011 x86_64 Six-Core AMD Opteron(tm) Processor 2425 HE AuthenticAMD GNU/Linux

After starting open-vm-tools-2011.04.25-402641 the information in vCenter changes to "Other (32-bit)", attached screenshot (vmware-tools_started.png) which is obviously wrong.

VMware vSphere vCenter Client/Server is Version 4.1.0 Build 345043.

Discussion

  • Timo Gurr

    Timo Gurr - 2011-05-05

    vmware-tools_stopped.png

     
  • Timo Gurr

    Timo Gurr - 2011-05-05

    vmware-tools_started.png

     
  • Greg Emerick

    Greg Emerick - 2011-07-19

    I'm seeing this on Fedora 9 as well with open-vm-tools-2011.06.27-437995. Anything I can provide to assist? vSphere 4.1.0 Build 260247

     
  • Anonymous

    Anonymous - 2011-07-20

    Tools will generally use the output of "lsb_release -sd" to figure out the OS information. Fedora 9 should have that; what's the output of that command in your VM?

    If lsb_release is not available, then that's probably the cause.

     
  • Greg Emerick

    Greg Emerick - 2011-07-21

    lsb_release -sd returns: "Fedora release 9 (Sulphur)"

    I can see where that would get me Other (32-bit).

    Could you give me some idea of what tools are looking for in the response to indicate 64 bit? If I have to, I'll make this return what's necessary, since it's not used by our appliance for any other purpose.

    Also, would there be any side-effects to just allowing this to happen? The vmx file is still set to Other (64 bit).

     
  • Anonymous

    Anonymous - 2011-07-21

    In general, the OS would be reported as "Other (32-bit)" if the code completely failed to recognize that output. The interesting bit is that we do have a check for Fedora in the code, and it should recognize that output. I'll need to try it out to see what's going on.

    As for what if affects, there are a couple of vCenter features that will look at that value and decide whether they'll run against the VM or not. But I think those features do not support Fedora anyway, so you should be safe. In your case it's probably just a cosmetic thing.

     
  • Anonymous

    Anonymous - 2011-07-21

    I just tried with Fedora 8 (which has a very similar output of "lsb_release -sd") and the code seems to detect Fedora correctly. I'll close this as an issue with the vCenter / VI client code (they may be missing the mapping to show Fedora correctly), but I shouldn't worry much because this is mostly a cosmetic issue.

     
  • Anonymous

    Anonymous - 2011-07-21
    • status: open --> closed
     
  • Timo Gurr

    Timo Gurr - 2011-07-22

    So what's the deal on Gentoo or (m)any other "Other" Linux distributions?

    # lsb_release -sd
    "Gentoo Base System release 2.0.3"

    Issue on the open virtual machine tools side, or time to open a vmware bug?

     
  • Timo Gurr

    Timo Gurr - 2011-07-22
    • status: closed --> open
     
  • Greg Emerick

    Greg Emerick - 2011-07-22

    I tried hardcoding lsb_release to return "Fedora release 8 (Werewolf)" and my image is still 32 bit. I've been looking at the code where lsb_release is called and it would seem that uname is also involved. Although command line uname appears to return very similar information as what I found online for FC 8. I sure don't see any issues with it.

     
  • Anonymous

    Anonymous - 2011-07-22

    Gentoo and many "Other" distros will probably show as their original configured OS. There's no point in adding every single distro out there to our UI, when only a few of them require tweaks at the hypervisor level. (I'd argue that the current list is already too long, but that's a separate discusssion.)

    As for the Fedora problem; unless you add some log statements to GuestInfoGather() in services/plugins/guestInfo/guestInfoServer.c (after the call to Hostinfo_GetOSName()), there's not much I can figure out from logs or anything else, because we don't log the raw data anywhere. "osName" and "osNameFull" are the strings provided to the UI to identify the OS.

     
  • Anonymous

    Anonymous - 2011-10-31
    • status: open --> closed-wont-fix
     
  • Anonymous

    Anonymous - 2011-10-31

    I'll close this as "won't fix" because (i) we consider this a bug in our UI and we have a bug tracking that internally, and (ii) there's not enough information to root cause the underlying issue in the bug.

     

Log in to post a comment.