Le Vendredi, 12 Décembre 2008 18:27:52 +0100, Unknown <vincent-perr...@club-internet.fr> a écrit :
Hi, > Each time I run a process in the UML machine, I get one more linux > process on the host, but the vsz and rss are always 0, the %cpu can > vary when the process works. If I look at the quoted ps output below, it looks like it got garbled. The thing is, I was on my way to write a little script that would verify the VSZ and RSS values periodically of about 20 UMLs and when if any rise in value then I should take a look at the host. I was thinking of doing that because on one machine these values always stays at 0, day in day out. When one rise, it is in parallel with 100% host CPU use by one of these processes. But, I can't do that. Au contraire, on another host these VSZ and RSS are constantly in the non-zero zone w/o any host CPU use. So it's not a valid test. I guess this is not a good approach after all. If I want to check CPU consumption on the host, then I should simply check that and not any derivative. Nevertheless, I'm curious about the UMLs VSZ and RSS values in correlation to their state. Also, why is there so many processes per UML ? On another subject, I quickly tried the uml_switch of clownix. I know this is not fair, because I did this very quickly, w-o even taking a longer look. But I had an error pointing to a line of code. I'll give you the actual error msg later. The regular uml_switch outputs a lot of: "send_sock sending to fd 4 Resource temporarily unavailable" Also, I've noticed that it takes time for the uml_switch to react when a ping is made for the first time after brining up a UML. I've noticed that when launching the 10th UML, from there to 20 UMLs. There must be a way to know what this fd 4 is, but I don't know how presently. None of the UMLs have disconnected during the brining up of 20 UMLs (all connected to the same switch). That also makes me think that perhaps starting 20 or more UMLs in a single command and doing ping tests as soon as they're up might not be a good idea. Some delay is perhaps necessary. But then, if you start 30 UMLs with 30 seconds delay in-between them then it becomes tedious. Salut, Al > Le jeudi 11 décembre 2008 à 20:44 -0500, lanas a écrit : > > Le Thu, 11 Dec 2008 22:06:14 +0100, > > Unknown <vincent-perr...@club-internet.fr> a écrit : > > > > Hi Vincent, > > > > > About the 100% cpu in uml, I tried to get it by playing with host > > > time (should never do that, ntp should always slide time slowly > > > to its target) I set the date a few seconds in the past: > > > > Now I'm more puzzled. For some reason I do not understand, I got > > entirely different VSZ and RSS values on another machine: > > > > USER PID %CPU %MEM VSZ RSS TTY STAT START TIME > > COMMAND user 28147 0.0 0.0 656 616 pts/6 T+ 08:50 > > 0:00 [linux] user 28287 0.0 0.0 180 164 pts/6 T+ > > 08:50 0:00 [linux] user 28823 0.0 0.0 600 600 pts/6 > > T+ 08:50 0:00 [linux] user 28833 0.0 0.0 384 384 > > pts/6 T+ 08:50 0:00 [linux] user 28853 0.0 0.0 1176 > > 1024 pts/6 T+ 08:50 0:00 [linux] user 28877 0.0 0.0 > > 472 472 pts/6 T+ 08:50 0:00 [linux] user 28879 0.0 > > 0.0 1204 1004 pts/6 T+ 08:50 0:00 [linux] user 28881 > > 0.0 0.1 1600 1408 pts/6 T+ 08:50 0:00 [linux] > > > > Whereas the previous (on another machine) were: > > > > USER PID %CPU %MEM VSZ RSS TTY STAT START TIME > > COMMAND user 8515 0.0 0.0 0 0 pts/6 T+ 09:11 > > 0:00 [linux] user 8655 0.0 0.0 0 0 pts/6 T+ > > 09:11 0:00 [linux] user 8889 0.0 0.0 0 0 pts/6 > > T+ 09:11 0:00 [linux] user 9239 0.0 0.0 0 0 > > pts/6 T+ 09:11 0:00 [linux] user 9255 0.0 0.0 > > 0 0 pts/6 T+ 09:11 0:00 [linux] user 9265 0.0 > > 0.0 0 0 pts/6 T+ 09:11 0:00 [linux] user 9285 > > 0.0 0.0 0 0 pts/6 T+ 09:11 0:00 [linux] user > > 9299 0.0 0.0 0 0 pts/6 T+ 09:11 0:00 [linux] > > > > These are idle linux systems, with extremely low CPU use in each. > > > > I was going to monitor the 0 VSZ and RSS values on the host to > > determine if the UMLs were in good shape, but now I cannot do that > > since it can be that UMLs are in good shape with other values. > > > > My current questions are: > > > > Why is there up to 8 of these [linux] processes per running UML ? > > > > And why are the VSZ and RSS values not uniform across machines ? ------------------------------------------------------------------------------ SF.Net email is Sponsored by MIX09, March 18-20, 2009 in Las Vegas, Nevada. The future of the web can't happen without you. Join us at MIX09 to help pave the way to the Next Web now. Learn more and register at http://ad.doubleclick.net/clk;208669438;13503038;i?http://2009.visitmix.com/ _______________________________________________ User-mode-linux-user mailing list User-mode-linux-user@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/user-mode-linux-user
