From: Simon N. <si...@pl...> - 2003-12-04 17:11:40
|
All, I run a couple of UML systems on one host (redhat 7.3 - yes I know, but its what the provider of the box installed on it). The UML systems are Debian (3.0r0.ext2 from the sourceforge UML homepage) and they communicate to the outside world via TAP devices. One of the systems has the following spec: 4 gigs drive, 128meg ram (limited at command line), external IP, runs apache, bind, imapd, pop3d, exim,proftpd. The apache server is heavily used, and the process ID numbers easily get up into the 3xxxx range.. The wierd thing is sometimes (when the PID's are high) the UML will start chewing up processor time on the host (the UML will show load avarges that look normal, but the host will show them looking a bit too high).. The UML will also behave strange when a 'ps ax' is done. The first 20 'ps ax' lines will come up, then the ssh session will lock solid. no ctrl-c , nothing. I can start a new session but cannot find the PId number to kill to get my original session back (ps ax will lock the new session also). ANy kind of PID interrogation program locks the session (including top). The UML will not reboot either (reboot at command line) - it simply broadcasts to all its intention to reboot then does nothing. After the reboot command, 'ps ax' on the host machine shows init trying to change runlevels, but nothing happens. the UML machine carries on in the same runlevel, running the same programs and doing it happily. I could leave the box in that state and it would go on untill i forcibly kill all the linux processes on the host machine. Has anyone seen this kind of behaviour/have any idea what could cause it/how I can stop it? I thought id fire this into the users lists before the dev list to get some insight.. Kind Regards, Simon Newton, Production & Support --- mBlox.com 11-13 Charterhouse Buildings London EC1M 7AP United Kingdom Tel: +44 (0)20 8432 1260 Helpdesk: +44 (0)20 8432 1285 Fax: +44 (0)20 8432 1290 E: Sim...@mb... --- |