Hello, Some of our Samba servers sometimes completely crash with this message on the console: "VM killed process smbd". If this occurs, we are obliged to power off our server and restart the server. You can no longer login nor can you enter the combination "Ctrl-alt-del". We use Red Hat Linux 6.2 and Samba 2.0.7. Does anybody have an idea what this message means and how to avoid these situations? I have no idea what VM means... If you have any information, feel free to contact me. Kind regards, Werner Maes KULeuven
Sounds oddly like a memory problem, but not necessarily. Unfortunately, you will probably need to spend more time than you want watching the server to see what is happening. VM is probably "Virtual Machine". Use something like "top" to monitor memory usage and the apps that use the most memory. Use "df" to watch hard disk usage. If you are running X windows all the time, change your default runlevel to 3 in the /etc/inittab file, so you just have a terminal. You can always type "startx" after you've logged on. After you start X, Red Hat has a number of visual tools to monitor memory and swap usage, xosview is my favorite, but xsysinfo is also good. (By the way, have you tried to use CTRL-ALT-F1 to switch to a terminal to log in, after X has crashed?) If you have a monolithic volume (mount shows /dev/hda1 /, and a swap file and that's it) then you may be running out of hard drive space. Seperate out /var and /home onto separate volumes, that way if they fill up, it won't lock the server. (Normally sending a large print job which is spooled to /var is the culprit.) Also you may want to look for core files to see if there is a memory dump. This will tell you the last things that the server was doing before the crash. Make sure you have enough swap space to hold your entire memory dump. If necessary, I think Linux will let you use multiple swap partitions. Also see if you have a run away process. Use "ps ax" to keep an eye on the processes throughout the day. I print them once an hour to trouble shoot this. See if any unusual ones keep multiplying or if there are a large number of zombie processes. Hope this helps. If not, give some more details and about the server and samba configuration. -----Original Message----- From: Werner Maes [mailto:werner.maes@cc.kuleuven.ac.be] Sent: Wednesday, November 08, 2000 11:36 AM To: samba@samba.org Subject: Samba server CRASH Hello, Some of our Samba servers sometimes completely crash with this message on the console: "VM killed process smbd". If this occurs, we are obliged to power off our server and restart the server. You can no longer login nor can you enter the combination "Ctrl-alt-del". We use Red Hat Linux 6.2 and Samba 2.0.7. Does anybody have an idea what this message means and how to avoid these situations? I have no idea what VM means... If you have any information, feel free to contact me. Kind regards, Werner Maes KULeuven
Thanks for your reply. On this server, X-windows has not been installed. We have two partitions. /dev/sda1 which is mounted on / /dev/sda6 which is mounted on /home (the partition with all the file shares). There's still plenty of space left on both partitions. We have 128 Mb swap space configured (our server has 128 Mb RAM). If I analyse with "top" I see that most of the memory is taken by the smbd-processes. How can I detect if there's a memory dump and where to find it these core files? Samba-configuration: This server acts as a file-server for 60 clients and has 3 shares. It is configured to authenticate against a password server. Kind regards, Werner Maes KULeuven "Roman, James (J.D.)" wrote:> Sounds oddly like a memory problem, but not necessarily. Unfortunately, you > will probably need to spend more time than you want watching the server to > see what is happening. VM is probably "Virtual Machine". Use something > like "top" to monitor memory usage and the apps that use the most memory. > Use "df" to watch hard disk usage. > > If you are running X windows all the time, change your default runlevel to 3 > in the /etc/inittab file, so you just have a terminal. You can always type > "startx" after you've logged on. After you start X, Red Hat has a number of > visual tools to monitor memory and swap usage, xosview is my favorite, but > xsysinfo is also good. (By the way, have you tried to use CTRL-ALT-F1 to > switch to a terminal to log in, after X has crashed?) > > If you have a monolithic volume (mount shows /dev/hda1 /, and a swap file > and that's it) then you may be running out of hard drive space. Seperate out > /var and /home onto separate volumes, that way if they fill up, it won't > lock the server. (Normally sending a large print job which is spooled to > /var is the culprit.) > > Also you may want to look for core files to see if there is a memory dump. > This will tell you the last things that the server was doing before the > crash. Make sure you have enough swap space to hold your entire memory > dump. If necessary, I think Linux will let you use multiple swap > partitions. > > Also see if you have a run away process. Use "ps ax" to keep an eye on the > processes throughout the day. I print them once an hour to trouble shoot > this. See if any unusual ones keep multiplying or if there are a large > number of zombie processes. > > Hope this helps. If not, give some more details and about the server and > samba configuration.