Red Hat Bugzilla – Bug 11313
TCP/IP based Pings get MAJOR packetloss outbound but not inbound.
Last modified: 2008-05-01 11:37:55 EDT
I have been recieving many calls about 6.2 having a problem with outbound pings and tcp/ip based communications suffering from 75%+ packetloss. I have not been able to reproduce the problem but I have had about 6+ calls in the last week. Example. Client has 6.1 tcp/ip transmissions work fine. no packetloss in or outbound. He upgrades to 6.2 is now getting 75%+ packetloss on ALL outbound traffic that is TCP/IP based yet inbound is fine. Pings inbound show up with no packetloss. client then does a fresh install and still has the same problem. client then takes off 6.2 and puts 6.1 back on. Works fine. NICS that have been called about are all tier1. -Matt
I can confirm this on my end. Updated customer from 6.0 to 6.2 Cannot ping anything outside of the box (99%-100% packet loss) Everyone can ping in with no problem. Samba even works, for while anyway. The MAJOR problem I am having is this install relies heavily on being able to use Samba. In less than 24 hours the entire network on the RH6.2 box dies, no in or out traffic. The system needs to be rebooted several times before it will work again. I have nuked and reinstalled 6.2, I was using a tulip based NIC, changed to a 3c90x but get the same result. The motherboard was changed at the time of upgrade, it is a ASUS P3V133 with the VIA 133MHz bridge. I have not reverted back to the old hardware to test for compatability issues. Are your clients using similar hardware? I hope this gets fixed soon. In the mean time I will be using 6.1 Robert
I think I found a way to fix this. I was doing more searching in bugzilla and found: http://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=10436 I tweaked the values in resolv.conf, seems to be working so far. Robert
If the fix was in /etc/resolv.conf, then the problem was the (new) necessity to add -n to ping when DNS is FUBAR.