Unable to query the live linux source machine

[alanrenouf]

I'm trying to convert a RHEL4 server and am getting 'Unable to query the live linux source machine' after I push Next on the credentials screen.

I can use putty from this server to logon as the root account. Any Ideas ?

The following is in the log files:

Testing client connection to agent

Reusing existing agent connection

Cannot query source HW info: converter.fault.SysinfoQueryLinuxFault

Testing client connection to agent

Reusing existing agent connection

Testing client connection to agent

Reusing existing agent connection

Cannot query source HW info: converter.fault.SysinfoQueryLinuxFault

Cached agent connection is for a different source, recreating connection

Creating SOAP stub adapter over pipe:
.\pipe\vmware-converter-agent-soap

Impersonating hitsar

Testing client connection to agent

Reusing existing agent connection

Cannot query source HW info: converter.fault.SysinfoQueryLinuxFault

Cached agent connection is for a different source, recreating connection

Creating SOAP stub adapter over pipe:
.\pipe\vmware-converter-agent-soap

Impersonating hitsar

Testing client connection to agent

Reusing existing agent connection

Cannot query source HW info: converter.fault.SysinfoQueryLinuxFault

[Texiwill]

Hello,

Cannot query source HW info: converter.fault.SysinfoQueryLinuxFault

Testing client connection to agent

Reusing existing agent connection

Two things here. One is that you can not get the info from the HW and you are reusing an existing agent connection. Was there ever a initial agent connection?

I would also look at logfiles on the Linux host.

Best regards,
Edward L. Haletky
VMware Communities User Moderator, VMware vExpert 2009
====
Author of the book 'VMWare ESX Server in the Enterprise: Planning and Securing Virtualization Servers', Copyright 2008 Pearson Education.
Blue Gears and SearchVMware Pro Blogs -- Top Virtualization Security Links -- Virtualization Security Round Table Podcast

[alanrenouf]

Checked the log files, all looks fine, the admin said he had had funnies with this box before and installing the backup agent, he said it had cpanell on it, Unfortunately I know nothing about this software.

Apparently it changes the OS a fair bit to make it more secure.

[vmweathers]

I'm not sure what "the backup agent" means. Nor am I familiar with "cpanell". Do you mean "[cPanel|http://en.wikipedia.org/wiki/CPanel]"?

Please post the Converter logs (File > Export logs...) Or if you are uncomfortable with that, send me a private mail and I'll reply with an email address you can send them to. Also please tell me what the server name/IP is so I can know what to look for in the logs.

(If your question has been resolved please mark the answers as "Helpful" or "Correct".)

[Texiwill]

Hello,

What commands is Converter trying to issue, it could be that one of those commands no longer exists on this host, is in the improper place or has been disabled in some fashion.

Best regards,
Edward L. Haletky
VMware Communities User Moderator, VMware vExpert 2009
====
Author of the book 'VMWare ESX Server in the Enterprise: Planning and Securing Virtualization Servers', Copyright 2008 Pearson Education.
Blue Gears and SearchVMware Pro Blogs -- Top Virtualization Security Links -- Virtualization Security Round Table Podcast

[vmweathers]

From the logs you sent to me:

cmdArgs[6]: mkdir /tmp/.vmware-sysinfo-ewyyifgmjuanfwas/

Converter Agent SysinfoQuery made temp directory; result: stdin: is not a tty

This might imply that cpanel has set "mesg y" in /etc/bashrc (or the equivalent): http://shreevatsa.wordpress.com/2006/03/16/stderr-is-not-a-tty-where-are-you/

This probably won't cause a problem in-and-of-itself though, since it is in the beginning of the cmd output and we'll ignore it.

cmdArgs[6]: cd /tmp/.vmware-sysinfo-ewyyifgmjuanfwas/; ./vmware-sysinfo-lin32.sh

\[Converter Agent SysinfoQuery] while trying to execute "C:\Program Files\VMware\VMware vCenter Converter Standalone\plink.exe -noprompt -stdin -nokeycheck -P 22 root@XXXXX cd /tmp/.vmware-sysinfo-ewyyifgmjuanfwas/; ./vmware-sysinfo-lin32.sh" received error code (126) with result: stdin: is not a tty

bash: ./vmware-sysinfo-lin32.sh: /bin/sh: bad interpreter: Permission denied

So that's the problem... for some reason /bin/sh isn't executable. Maybe the /tmp filesystem is mounted with "noexec" set? Can you remount it without that flag if so?

(If your question has been resolved please mark the answers as "Helpful" or "Correct".)

[bracewell]

I am seeing the same error as described in the first post. I am attaching parts of the vmware-converter-agent.log. I looked on the RHEL box for "converter" log files but did not find any. Can someone tell me the location they are stored?

[vmweathers]

There are no converter logs left on the source machine after Converter is done querying. (Assuming the Converter server or GUI are not running on the source.) You can see in your logs that the vmware-sysinfo binary crashed on your machine (segfault).

I've attached versions of the sysinfo binary that will hopefully emit a traceback if the crash occurs, can you please try using this one manually on the target? (i.e., unpack it and run ./vmware-sysinfo-lin32.sh) It may not work in the whole converter process so I do not recommend replacing the existing sysinfo tarball with this one.

(If your question has been resolved please mark the answers as "Helpful" or "Correct".)

[bracewell]

I successfully executed the vmware-sysinfo.sh and it created vmware-sysinfo.log. The log data looked error free. I have included the log. What is the next step? Many thanks!!

[nwaldon]

according to the sysinfo log your kernel is xen enabled. THe release notes mentioned that xen kernel are not supported by default. It also mentions a work around:

You cannot install Converter Standalone 4.0 on a Linux machine running with a Xen kernel

A check in Linux tar installer prevents Converter Standalone 4.0 from being installed on a Xen kernel.Workaround:

1. Unpack the Linux installer tar package.

2. Inside the unpacked installer, open the installer script<code># vmware-installer.pl</code># using a text editor

3. In<code># vmware-installer.pl</code># , search for<code># /proc/xen</code># to find the check for Xen kernel. The check denies installation of VMware products unless the product is<code># tools</code># . Comment out this check to disable it.

4. Save the changes to<code># vmware-installer.pl</code># and start it to install Converter Standalone.

http://www.vmware.com/support/converter/doc/releasenotes_conv40.html

let us know if that helps.

[vmweathers]

Hmmph... that is what I was afraid of. My changes were not meant to "fix" anything, they were solely meant to emit a traceback if the crash occurred. But unfortunately those changes altered the program just enough to somehow prevent this crash from happening anymore. Can you take the original vmware-sysinfo-lin32.tar.gz, unpack it and run it? I need a core file from the run, so you might need to enable coredumps first (ulimit -c unlimited). Also please look in /var/logs/ for any references to tcrash/segfault/oops at the time of the crash. Also I'll build a version of the original binary that has symbols and ask the same thing later, if the coredump doesn't produce sensical output.

(If your question has been resolved please mark the answers as "Helpful" or "Correct".)

[vmweathers]

Note that the xen kernel in-and-of-itself should not be responsible for the crash, but it is possible.

Also please note that you seem to be using a beta version of RHEL. From searching the web "RHEL 4.92 (Tikanga)" seems to imply that this is RHEL 5 beta 2.

That is not supported by Converter Linux P2V, however I would like to get to the bottom of these crashes. Interestingly, I've seen 3 crashes: 1 on mandriva, 1 on debian, and 1 on RHEL 5 beta 2 (yours). None of these are supported platforms, so there is some correlation there. Hopefully I'll be able to find a RHEL 5 beta 2 ISO and install this myself and hopefully reproduce the crash.

(If your question has been resolved please mark the answers as "Helpful" or "Correct".)

[bracewell]

Here is the core file.

[alanrenouf]

From the logs you sent to me:

cmdArgs[6]: mkdir /tmp/.vmware-sysinfo-ewyyifgmjuanfwas/

Converter Agent SysinfoQuery made temp directory; result: stdin: is not a tty

This might imply that cpanel has set "mesg y" in /etc/bashrc (or the equivalent): http://shreevatsa.wordpress.com/2006/03/16/stderr-is-not-a-tty-where-are-you/

This probably won't cause a problem in-and-of-itself though, since it is in the beginning of the cmd output and we'll ignore it.

cmdArgs[6]: cd /tmp/.vmware-sysinfo-ewyyifgmjuanfwas/; ./vmware-sysinfo-lin32.sh

\[Converter Agent SysinfoQuery] while trying to execute "C:\Program Files\VMware\VMware vCenter Converter Standalone\plink.exe -noprompt -stdin -nokeycheck -P 22 root@XXXXX cd /tmp/.vmware-sysinfo-ewyyifgmjuanfwas/; ./vmware-sysinfo-lin32.sh" received error code (126) with result: stdin: is not a tty

bash: ./vmware-sysinfo-lin32.sh: /bin/sh: bad interpreter: Permission denied

So that's the problem... for some reason /bin/sh isn't executable. Maybe the /tmp filesystem is mounted with "noexec" set? Can you remount it without that flag if so?

(If your question has been resolved please mark the answers as "Helpful" or "Correct".)

Thanks for your help so far, I am now getting a little further and failing with an error copying the /boot partition. I have mailed you my latest logs, I hope that is ok.

[vmweathers]

hi Alan. The problem is that the source hostname is not being resolved by DNS in the helper. You can see this in the UnixP2VVolumeCloneFailedEvent in the agent and server logs:

ssh: Could not resolve hostname XXXmaskedoutXXX: Name or service not known

BTW, since this is not nicely exposed to the UI right now, I'm going to make a fix that will make the user visible error be: Unable to connect to 'hostname'

I see that you did set the DNS Servers in the HelperVmNetwork static IP configuration plugin on the Converter wizard's 3rd page. However you did not set any DNS suffixes. You'll need to set those if you want the unqualified hostname to be resolved.

Or you could use the FQDN of the source host.

Or you could just use the source host's IP in the 1st page of the wizard.

(If your question has been resolved please mark the answers as "Helpful" or "Correct".)

[alanrenouf]

Oh man, I cant believe I did that, I thought it had got past that bit as I was able to select the partitions to copy/expand and every other option. I always forget that everything has to talk to everything.

Perhaps you could get it to say "Helper VM Unable to connect to 'hostname' "

I have more scheduled downtime tomorrow morning so will let you know how it goes.

Thanks again for your help.

Alan

[vmweathers]

Thanks for the suggestion.

Some more tips that will hopefully ensure tomorrow's Linux P2V goes as smoothly as possible:

You should probably make sure there are no large sparse files on the source. See:

Also there are 2 other VMware VC add-ons that can wreak havok on the HelperVM:

make sure you do not have VMware HA enabled (if you are using VC 2.5u3 and ESX 3.5u3). See:

make sure you are not running VMware Update Manager (VUM) version 1.0 or 1.0u1 (i.e., VUM 1.0u2+ is ok).

(If your question has been resolved please mark the answers as "Helpful" or "Correct".)

[markokobal]

Hi,

I had the same error reporting from Converter, but the problem has bee elsewere. After one day of struggling for solution I have found out that problem was in network interfaces naming. I had names like "eth-something", "eth-somethingelse" and the problem was the "-" char in the interfaces names! There should be only a-z chars in interfaces names, no dots "." or dashes "-", etc.! Seems like I bug... Anyway, when I changed the names to standard names "eth0", "eth1"... the Converter was able to connect and did the conversion. The source machine was Debian5, destination ESXi4.

--

Kind regards, Marko.