On Wed, Jul 13, 2005 at 12:06:31PM -0700, Jay Truesdale wrote: > One of my servers started having problems after it had been stable for about > 10 > months. Communications in/out of the box would just stop or the machine would > crash. We were able to log in at the console. A ping command returned the > error > message "no buffer space available." I searched the list archives and found a > few references to this error message along with some suggestions, but nothing > that ever said what fixed the problem. > > We may have fixed this problem on our box so here is what we did in the hopes > that it may help someone else if they run into this problem. > > Our box was running OpenBSD 3.6 with most of the patches installed. After a > crash one day and communications getting stuck the next (/etc/netstart got the > network back running again, btw), I installed all patches that were TCP > related > or could have been somehow related to the problem. I rebuilt the kernel > (generic) and rebooted. > > A few days later the machine hung up again. There was an on-board NIC and a > PCI > NIC installed in the box and they were sharind an interrupt which one of the > postings I read said was a bad idea. Disabled the on-board sound and on-board > NIC in the BIOS so that there were no shared interrupts. > > The next day the machine hung up again and /etc/netstart got the network > working again. After some discussion we removed the PCI NIC card and enabled > the on-board NIC card and changed hostname.xxx to match the on-board nic card. > That was on July 5, 2005 and the system has been stable for the last ten days.
It would be nice if you could let everyone know what NICs were in use and provide a dmesg from the system.
