Date: Wed, 4 Jun 2003 22:26:18 +0000 From: Daniela <dgw@liwest.at> To: "Erik Paulsen Skaalerud" <erik@pentadon.com>, <stable@freebsd.org> Subject: Re: Server overloaded? Or is it a bug? Message-ID: <200306042226.18706.dgw@liwest.at> In-Reply-To: <001c01c32a1a$da2092f0$0a00000a@eps> References: <001c01c32a1a$da2092f0$0a00000a@eps>
next in thread | previous in thread | raw e-mail | index | archive | help
On Tuesday 03 June 2003 21:55, Erik Paulsen Skaalerud wrote: > > > Did you try looking at its console? > > > It probably have lots of pretty messages of whats going on.. > > > > Can't look at the console. It hangs completely. > > Hm. If its running X on the console, have you tried forcing it to a vty? > (ie, alt+f2 or similar). If that doesnt work, I really dont know what to > say but to reboot it and see if you are lucky, -maybe- "dmesg -a" have some > errors in it from the failure. Or maybe /var/log/messages. It depends on > what the failure was really. I just rebooted, but there are no messages. About one minute before the crash, when I wasn't able to run any more programs, I looked at the console (nothing special there), tried to login on ttyv6 (could enter the username, but no password prompt) and went back to X. There I tried to find out what has happened. The cd command was working as expected, but top and ps (for example) did just nothing. When I opened a new terminal session, it showed up, but I didn't get a prompt. > > > > I don't have much running, just KDE with about 10 > > > > programs on each > > > > > > of the 16 desktops, and a few background processes. This > > > > seems much, > > > > > > but I often have much more stuff running, and it is not even slow. > > > > > > > > It does respond when I ping it, but won't let me in over SSH. The > > > > processor doesn't sound busy, so I suspect that the scheduler has > > > > gone away, or there is some bug in the kernel, or some > > > > system table > > > > > > is too small. > > > > > > Hm. What kind of sound does a busy CPU make? :-) > > > > > > > How do I find out what the problem is? (Never had any > > > > before.) > > > > > > Yeah, attach a console and find out whats going on. > > > > How can I do that? > > It doesn't respond to ctrl+alt+backspace and ctrl+alt+del, I > > can still move > > the mouse, but I can't click anything. > > See above > > >(I known I shouldn't run X11 on a server.) > > You're right, the console of a server should not run X11 :( > If you insist on running X11 on the console, maybe you should try to always > keep an xterm open with console messages? (see /etc/syslog.conf) X11 doesn't run on the console, it runs on ttyvb. And it doesn't run always. I only need it when I have to see graphics (I'm a command-line freak :-)). > > > > FreeBSD has always managed the highest loads, even > > > > on normal PC hardware. Is it possible to bring the server back > > > > without rebooting? I would lose a lot of unsaved data if I had to > > > > reboot. I'm running 4.8-STABLE. > > > > > > When a server responds to ping, but ssh times out etc it > > > > often relates > > > > > to hdd problems. Atleast in my experiences it has been so > > > > (dead disk). > > > > > And if it is hdd problems, and FreeBSD couldnt get the disk > > > > to wake up > > > > > again, you probably already have lost data. > > > > That could be the cause. One minute before the crash I could > > still enter shell > > builtin commands, but external commands did just nothing. > > > > Is it possible that the process table is full (I had more > > than 450 processes > > last time I ran top) or I ran out of memory (512M RAM + 1024M swap)? > > Well. Lets say that all the users use the same application. And all the > users got the same bug with the application, wich made it forkbomb and > consume large amounts of memory. 450 user-procs on the runaway could make > this happen, but its very unlikely. FreeBSD is good at locking down things > like this. But then again, if you run out of swap you're busted. I guess it's something with the processes. I'm pretty sure the kernel was still alive. The maximum number of processes is (maxusers * 20) + 18, right?
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?200306042226.18706.dgw>