I have NOT opened an SR on this, because (1) I can workaround it by having static routes that prevent my routers from ever having to redirect a packet, (2) almost NOBODY has the problem and it may be something strange about my network, and (3) Oracle Linux/VM support frustrates the hell out of me with their slowness and general incompetence.
YES, 2.6.32 works fine, and 2.6.39 does NOT.
We can, however, work around it by using static routes. The problem has something to do with routing "going bad". And I'm pretty sure the problem begins when a router responds with a 'redirect' back to the OEL server, when the server sends a packet to the "wrong" router.
Oh, one other workaround, for me, is to flush the route cache often. When the machine is hosed up, I can SSH into it from another souce, and run "ip route flush cache" (or whatever the syntax is) and that instantly fixes the problem. If you were to run that command in a cron script every minute, it might keep you going until the problem gets fixed.
I'm running 2.6.39-200.1.9.el5uek in Oracle MV that might be why I've never seen the error. Where did 2.6.39 get introduced in Oracle VM? or is it just VM guests? I do run a couple OEL6 VM guests in test but that's it. Never seen the issue there and the kernel is 2.6.39-400.17.1.el6uek.x86_64
Could it be related to / influenced with any of the following settings?
I had issues with multiple subnets on the same LAN, causing the linux arp cache to get polluted with arp entries from hosts on different subnets. That was solved by setting the following:
We had an SR for over 2 month now. After sending a lot of reports an information there are no findings from support.
We are closing this ticket now because of time and hardware resource problems and because we don't think that this SR can be successful.
If you disagree you should preform an escalation:
How To Escalate a Service Request (SR) with Oracle Support Services (Doc ID 199389.1)
have a good day,
My issues turned out to be some type of kernel corruption. I still don't know what caused it. I upgraded the kernel to the newest UEK, used DKMS to generate some non native drivers....and all my issues went away. I've never had the issue with Oracle VM but I can now see where it might happen. My issue was a Oracle Linux 5 native host... no xen.