Description of problem: I have Oracle 10gr2 installed on two nodes. They are clustered using OCFS2. The Kernel is running in 64bit mode with version 1.0.9 of OCFS2 module. When I issue a shutdown -r on node 1 both systems panic. It seems that any time a node drops out of the cluster the other node panics as well. In addition we have opened a ticket with Oracle. Version-Release number of selected component (if applicable): 2.6.9-22.0.1.ELsmp x86_64 How reproducible: Easily Steps to Reproduce: 1. Install Red Hat AS4 x86_64 on Intel Xeon 2. Install OCFS2(1.0.9) from Oracle 3. Install Oracle 10gr2. 4. Cluster the two nodes using gigabit interconnect. 5. Remove one node from cluster by rebooting, or dropping connection. Actual results: One or both nodes immediately panic. Expected results: The untouched node should remain running without causing a kernel panic. Additional info:
Created attachment 123455 [details] node one 1 error log
Created attachment 123456 [details] ndoe two error log
Upon further research it looks like this may be an Oracle o2cb configuration problem. But then again it isn't quite behaving like I expect the self fencing to behave... I'll dig further and see what I can find out.
ok. thanks. I did think that there was a chance this was an oracle issue based on the node 2 error log...I'm going to put this in NEEDINFO, pending an update from you. thanks.
I got this from Oracle's Bugzilla: Quite a few bugs have been filed on this issue. See bug#630. http://oss.oracle.com/bugzilla/show_bug.cgi?id=630 So this looks like a 'feature' and not really a bug. ;) I've modified the shutdown/reboot scripts so that the ocfs2 module is unmounted at the same time the network is stopped. This solved the problem for us. It's a little annoying that the default behavior of the latest ocfs2 rpm's is not quite right, but I've made it work.