So you’re sitting around, trying to get rid of some ridiculously old legacy hardware (and the oooold OS that’s on it) that you cannot find time/resources to upgrade. You think, “Oh I know, I’ll just virt it and deal with it later!”. Pretty reasonable, you’re a goddamned genius. So you install a copy of vmware converter and get to it.
Just for arguments sake (ok, and some wording to help folks like me that spent too long googling for an answer), let’s say that you are needing to do a p2v conversion of fedora core 3. You have loaded your windows machine with the latest VMWare Converter (5.0.1 as of this writing), but when you try to connect to the linux server it says “unable to query the live linux source machine”.
You then decide to try to outsmart it, so you find an older copy of the converter that runs on linux (I use 4.0.1), push it over and install it on the server. After a half hour of futzing around with DISPLAY commands and upgrading the SSH server on it, you finally manage to export your X session so that you can launch the vmware-converter-client. But then, sure enough, it responds with the same message.
So you google some more, and you find some terribly unhelpful VMWare articles on things that might go wrong (bad /tmp permissions, unable to execute scripts, etc) and realize none of those things are an issue. So what exactly is still going wrong?
Beats me. BUT if you’re lucky, it’s as simple an answer as it was for me. Do your ethernet interfaces have any unconventional names? For example, I had several interfaces on my old fedora server that contained dashes and dots in them. I moved those conf files from /etc/sysconfig/network-scripts/ to a different folder and restarted the server (for some reason just downing the interfaces wasn’t enough to make it work for me), and the converter was suddenly more than happy to p2v for me.
So there ya go. TL;DR — if your interface names are anything non standard (ie, not just eth0, etc), then try renaming or removing them, and THEN try the p2v conversion.