<html><div style='background-color:'><DIV class=RTE>Hi, just installed 10g RAC on ocfs2 with 2 nodes, on an RHEL AS 4 x86_64 server (4gb quad opteron).</DIV>
<DIV class=RTE> </DIV>
<DIV class=RTE>Everything seemed ok until the DBA started to build the database and do some heavy operations to it. Node 0 on the cluster kernel panics after this console message:</DIV>
<DIV class=RTE> </DIV>
<DIV class=RTE>(6,0):o2hb_write_timeout:164<BR>ERROR: Heartbeat write timeout to device dm-0 after 12001 milliseconds<BR>(6,0):o2hb_stop_all_regions:1673<BR>ERROR: stopping heartbeat on all active regions.<BR>Kernel panic - not syncing: ocfs2 is very sorry to be fencing this system by panicking</DIV>
<DIV class=RTE> </DIV>
<DIV class=RTE>on Node 1, this was on the console:</DIV>
<DIV class=RTE> </DIV>
<DIV class=RTE>(2585,1): o2net_set_nn_state:421<BR>no longer connected to node DC1ORA01 at 192.168.79.169:7777<BR>(32763,1):ocfs2_replay_journal:1123 Recovery node 0 from slot 0 on device (253,0)</DIV>
<DIV class=RTE> </DIV>
<DIV class=RTE>and Node 1's OS was barely responsive and wouldn't shutdown cleanly.</DIV>
<DIV class=RTE> </DIV>
<DIV class=RTE>The DBA said Oracle was creating numerous trace dumps due to I/O errors, especially during heavy load. Where do these errors point to? storage drivers? ocfs2 bugs? incompatibilities with 10g RAC and ocfs2? Where do I start here?</DIV>
<DIV class=RTE> </DIV>
<DIV class=RTE>Oh, device dm-0 is a standard logical volume made up of 3 physical volumes from a SAN array. I downloaded the 10g install disk images to it and installed just fine, so the storage appears to be working properly and does for other server environments.</DIV></div></html>