guest@stat.fsu.edu (a guest account) (08/06/89)
We have a VAX 11/780 running (so to speak) Ultrix 2.0. Periodically, it stops keeping time, which has all types of adverse effects. cron doesn't get run, update doesn't happen, any logins over the net get hung, init doesn't spawn new gettys after a shell terminates; trying to use shutdown or /etc/halt do not work because they too have timeouts (/etc/halt -q does work). This will go on for varying amounts of time: anywhere from 3 or 4 minutes up to 12 hours, then suddenly the clock will start ticking again. The only weirdness I can spot is that icmp_error() is getting called a lot (about once a second on average, with large peaks of 20+/second and then blank spots), with icps_oldshort getting incremented to about 80% of icps_error. This all started around July 17. I have checked all of /etc and /usr/lib; all of the binaries seem to be the same as that running in February. The frequency of seizures seems to be related to the amount of daemons running; if I turn accounting and quotas off, and just leave update, cron, inetd, sendmail (gotta have that mail), and elcsd running, lockups become very rare, and not too severe (once about every 3-4 days.) If I add lpd, the system will seize up quickly and often. This certainly has puzzled me... If anybody has seen anything like this, or has any ideas, please mail me at the below address. One of the first things I took off was news, so I am posting this courtesy of another machine, and I won't be able to read this newsgroup. If this problem seems to be of enough currency (not rare hardware weirdness), I will post a summary of the/any solution(s) I get. Thanks in advance, Randolph Langley langley@nu.cs.fsu.edu