On Sat, 18 Sep 1999 15:21:28 -0400 (EDT), William T Wilson wrote: >On Sat, 18 Sep 1999, Ralf G. R. Bergs wrote: > >> modules or overclocked CPUs.) Memory usage is permanently about 99%, swap >> usage only a few percent. But obviously processes are dying because they > >Then you are not running out of memory. The kernel likes to leave all the
Well, this is what I initially thought too. Once upon a time :) I learnt that under a system that uses virtual memory swap is as good to a process as RAM (apart from access time, of course.) Therefore in theory if I make my swap file large enough I should be fine, and no process should ever die from memory shortage. Right? Ok, I will substantially increase my swap memory on one machine from 64 MB to 128 MB (48 MB RAM) and check whether it makes a difference. >memory allocated. It uses it for a disk cache and just gives it to user >programs as necessary. You'll get "can't get a free page" messages on >console when you really start to run out. Normally this will happen when I can't see console output because the machine in question doesn't have a console. I only occasionally plug in a serial cable connected to a workstation in the same room (if I goofed really bad :-) >> I need a high availability of my machines, and having to constantly >> check and possibly restart services is not acceptable. :-( > >What is the problem with your inetd? It simply dies, as do several other services, like Apache, SNMP, Samba, etc. -- Sign the EU petition against SPAM: L I N U X .~. http://www.politik-digital.de/spam/ The Choice /V\ of a GNU /( )\ Generation ^^-^^
