Is it a dell 1550 by any chance? On Sun, 2005-04-24 at 10:43 -0400, Robert Sanders wrote: > Casey, > > We've been seeing issues like this for probably the last year. I was > never able to pinpoint it to any action. We implemented remote reboot > hardware and called it a day. > > Some of them had strange activity, but over a larger group of machines I > could never find a pattern to it. It almost seems as if it cannot spawn > any new processes. > > I can't help except to say your not alone. > > Rob > > Casey Allen Shobe - SeattleServer Mailing Lists wrote: > > Hey all, > > > > We're seeing occasional issues with a bunch of machines we have in a > > datacenter, most of which are currently running Gentoo. The machines will > > run solid and fine for days, weeks, even months, and then just lock up > > solid > > - the box still pings and an nmap scan shows all the normal ports open, but > > nothing responds on any port, nothing shows up in system logs, and the > > times > > we've had console access to a machine at the time, a login prompt would > > show > > up, but it would just hang if you tried to log in. > > > > This generally indicates hardware issues to me, but it has been happening > > across a wide array of both well-tested and new machines. In addition, it > > happens on machines that are running Red Hat 7.1 through 9.0 as well as > > Gentoo. The problem seems random, and there is almost always close to zero > > load on the machine when it locks up (only once were we presently using the > > machine, and it locked up while uncompressing a tar file). > > > > The Gentoo systems use the deadline I/O scheduler as it's deemed the most > > reliable, but this has shown up with the default anticipatory I/O scheduler > > as well. > > > > The only common factor seems to be that they are all plugged into a > > questionable HP Procurve switch that we've been contemplating replacing. > > Would that simply be wasting our time (I don't think a buggy switch should > > be > > able to lock up boxes...)? Any recommendations for what to investigate at > > this point? > > > > Cheers, >
-- [email protected] mailing list
