pvsnatd running wild

Sometimes pvsnatd acts up and eats all available CPU cycles. I have the following setup: Host Mac OS X Leopard with two guests, one WindowsXP and the other one RedHat Linux. Both guest use shared networking thus forming a virtual LAN. I have enabled some port forwarding in Parallel so that Windows XP can share drives with the outside world and RedHat can offer Oracle connectivity (there is Oracle running on RedHat). I do Windows specific client/server development against the Oracle server. The Oracle Server itself must access other servers in the outside world (outside means outside of my Mac). The connection to this outside word is direct when I am at the customers site but through OpenVPN when I am remote (the Mac is a MBP with 4GB RAM). On the other hand when I am at the customers I routinely connect again with OpenVPN from the Mac to my own office server. So most of the time I have that virtual LAN "inside" Parallels, a direct connection to the external LAN from the Mac and an OpenVPN connection so some remote LAN. In this scenario often pvsnatd starts to eat CPU. Only a reboot of the whole Mac can then remedy things. Is this something that is known? Any workarounds?

Thanks for listening and help

 

Does the problem occur if you turn off your port-forwarding?

 

Can't tell because the system is being used productively during the day when I am working onsite at a customer. The problem manifests itself only after a couple of hours, sometimes only after many many hours (I often leave the MBP running overnight for off-hours compiler runs etc). So it is not feasible to turn off port forwarding and wait for the issue (not) to happen. Without port forwarding I cannot to my work.

 

Ok, I'll think about it and tweak with a few things and see. It's good to know the timeframe of the problem happening, though.

 

Hi,
for me the pvsnatd acts the same way.
Actualy I don't have to wait a bit it eats up my resources right after booting the system. I don't need to start or use the Parallels not a little bit. During the wildeness it also have some small but distinct network traffic.
Maybe it is completelly unrealted but I was moving to this flat - in new country with new network - recently. Since this happened (2 weeks time) I did not start the Parallels here. Its been used in the previous place last time.

 

It just happened again (more info)

Just now it happened again. I have started a VPN connection from the Mac side to my office. About 30 minutes later I opened a WinXP and a RedHat virtual machine, both with shared networking. Everything fine. After only a good hour later all of a sudden pvsnat started to eat cpu cycles and at the same time syslogd also went crazy. With the very same setup the system run through during the last couple of days without any problem (no sleep, no shutdown, no quitting parallels, no shutting down VPN, no whatsoever). Only last night I shut down to take the computer home and bring it back in today. So the problem sometimes shows up only after several days of constant use and sometimes almost immediately.

system.log has the following entry:
Mar 13 08:59:39 gugus pvsnatd[226]: pvsnatd(226,0xa0402fa0) malloc: *** error for object 0x104190: double free\n*** set a breakpoint in malloc_error_break to debug
Mar 13 08:59:45: --- last message repeated 138 times ---

 

pvsnatd is usng almost 100% CPU.

I have Parallels 5584 on my Core2Duo MacBook Pro.
MacOS: 10.5.2
2GB of RAM

I am using Bootcamp, but this problem happens before I start Parallels after a reboot.
I am using port forwarding.

Problem could be related to:
I have a program called Wuala that is running on the Host OS, but listening for connections on the wrong interface.
Parallels is giving out IP's through DHCP on 10.211.55.1-254
Wuala is listening on 10.211.55.2
It also says that is can't connect to the internet until AFTER I kill pvsnatd by running sudo /Library/StartupItems/Parallels/pvsnatd stop

 

Totally bizzare. Have you tried configuring the port forwarding by the ipfw and natd daemons in OS X directly?