Dear all, I would like to request some hint from you to sort out an issue while evaluating HA-NFS-iSCSI on OpenSolaris 2009.06 + COMSTAR + OHAC (and so far did not found a successful answer neither on here nor google).
Let me summarize the system layout (all x86 hosts): - jbod1 host: to export via COMSTAR a bunch of local disks (i.e. one LUN per entire disk c0tNd0 where N = 1..8). Accessible via an IPMP group of two interfaces on different switches. - sanfe1/sanfe2 hosts: OHAC cluster with two NICs in IPMP group, two cluster interconnects and two vlans in IPMP group to access the jbod1 (these vlans are on top of the cluster interconnects). - In non cluster mode, the initiators (sanfe1/sanfe2) perfectly mount the disks from jbod1 and I can create zpools (with different raidz layouts and do stress tests via NFS mounts) The problem: - in cluster mode, sometimes creating the zpool with DID devices (/dev/did/dsk/dN or /dev/did/dsk/dNs2), sometimes accessing the zpool via NFS and creating load, the underlying host panics and not even completes the crash dump I tried changing parameters like MPxIO, fencing, but no luck. Could you please give some hints about how to dig deeper into the problem? To your knowledge, is there any incompatibility among OHAC, COMSTAR, OpenSolaris 2009.06? Thank you all, Carlos -- This message posted from opensolaris.org
