Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!uunet!stanford.edu!morrow.stanford.edu!pangea.Stanford.EDU!farrell From: farrell@pangea.Stanford.EDU (Phil Farrell) Newsgroups: comp.unix.ultrix Subject: Re: Cant get mbufs Keywords: mbufs Message-ID: <1991Jun3.173845.19071@morrow.stanford.edu> Date: 3 Jun 91 17:38:45 GMT References: Sender: news@morrow.stanford.edu (News Service) Distribution: comp Organization: Stanford Univ. Earth Sciences Lines: 25 In article rcmart@urc.tue.nl writes: > >Since a few days we have a very nasty problem on our DECsystem 5400, >which runs under Ultrix 4.0. >After a timeperiod of about 7 hours the system reports "cant get mbufs" >on the console and then the system hangs! "mbufs" are the memory areas used to queue incoming and outgoing ethernet packets. You can see how many are in use with "vmstat -K". The number allocated by the system is a function of the "physmem" and "maxusers" kernel configuration parameters. Perhaps your network load is simply too great for the values you have configured into your kernel. You could try increasing either or both of these parameters and building a new kernel. By the way, DEC lore (that is, what the support engineers tell you, not what the documentation says) is that the "physmem" parameter on RISC systems should not be set equal to the actual amount of physical memory, but rather to at least 125% of actual memory in order to properly size system tables, particularly the PTEs that map between physical and virtual memory. On the other hand, you may be seeing a pathological condition where mbufs are not being properly freed. I can't help you with that possibility and have my doubts that DEC can either. -Phil Farrell, Computer Systems Manager Stanford University School of Earth Sciences farrell@pangea.stanford.edu