Dealing with a runaway process in Slackware 15.0
by HicEtNunc from LinuxQuestions.org on (#6JYN9)
Once in a while (actually, twice in the last month) I get a runaway process in my Slackware 15.0 box. When that happens, the system becomes unusable: in fact, my XFCE desktop freezes up, in that the clock applet does not update and the mouse (and keyboard) become totally unresponsive. Not necessarily forever: after a few minutes (or more than just a few minutes) the system may become (somewhat) responsive again, at least for a while. Eventually, however, it is just thrashing, and I have to do a hard reboot on it. This is extremely irritating, not only because you never know what is going to happen after a hard reboot, but also because I lose all my current setup, which may consist of dozens of terminal emulators. Restoring this is painfully tedious.
What I would like is to be able to kill whatever runaway process is hogging all the resources - this process may sometimes be my doing (don't ask) or sometimes something at my browser (Chromium).
Any suggestions on how to deal with this? The obvious "don't start such processes" will not do - it is not clear a priori what process can become a runaway one, especially when it happens at the browser. Trying to kill it by hand will also not work, for obvious reasons. I guess that what I would be looking for is some mechanism such that when a process meets certain requirements (like e.g. sending the system load above some specific value, like, say, 100) or trying to allocate memory beyond a certain limit, the kernel would just kill said process immediately.
Does something like that exist?
What I would like is to be able to kill whatever runaway process is hogging all the resources - this process may sometimes be my doing (don't ask) or sometimes something at my browser (Chromium).
Any suggestions on how to deal with this? The obvious "don't start such processes" will not do - it is not clear a priori what process can become a runaway one, especially when it happens at the browser. Trying to kill it by hand will also not work, for obvious reasons. I guess that what I would be looking for is some mechanism such that when a process meets certain requirements (like e.g. sending the system load above some specific value, like, say, 100) or trying to allocate memory beyond a certain limit, the kernel would just kill said process immediately.
Does something like that exist?