dmr@csli.Stanford.EDU (Daniel M. Rosenberg) (02/27/90)
Our continuing sendmail problems, part 3... First, thanks to everyone who replied with useful suggestions. While I think none of them are the problem (we are runinng no sendmail.fc, and we are running the same sendmail.cf, resolv.conf and named configuration as every Cube at Stanford). But, my own digging around has led me to this hypothesis: our messages queue forever and net connections are disallowed because our load average is too "high." Well, it really isn't, but uptime gives 9.19pm up 7362 days, 5:19, 3 users, load average 1701650.75, 0.00, 1886352.00 And those excessive 1 and 15 minutes values never change. A quick look at our perfectly legal copy of the source reveals that uptime and sendmail get their load average information from the same place, /dev/kmem, and when sendmail sees a load average that high, it just queues messages. This is a feature. Problem is, we can neither seem to make our load average normal, nor can we fool sendmail with the x and X options to let it work anyway. So, anyone have experience with excessively high, excessively false load averages? Thanks for any tips! -- # Daniel M. Rosenberg // Stanford CSLI // Eat my opinions, not Stanford's. # dmr@csli.stanford.edu // decwrl!csli!dmr // dmr%csli@stanford.bitnet