Some comments, On 01/07/10 12:11, Michael W Lucas wrote: > I am trying to set up a NFS cluster as per the instructions at > http://docs.sun.com/app/docs/doc/820-7821. I've reached the point where I > export disks over iSCSI with COMSTAR, and the process fails. Being new to > OpenSolaris, I'm sure I'm missing something basic. > > I have two 2009.06 x86 servers, each with two public NICs (one to the mgmt > network, the other to the client network), and a third NIC connected via > crossover cable to the other server. Each has ha-cluster-full installed. > > I think I have disks exported. Here's one server exporting c10t0d0: > datastore1~;itadm list-target > TARGET NAME STATE SESSIONS > iqn.1986-03.com.sun:02:6230a9ff-67d0-e995-9e85-f009c17adab9 online 0 > datastore1~; >
The target needs a list of logical unit(LU) attached to it using a view. Did you create a view for that logical unit? The command "stmfadm add-view" would achieve this. In the document at, http://docs.sun.com/app/docs/doc/820-7821/gilon?a=view see example 3-2. > Here's the other server, where server 1's c10t0d0 is listed as the first > target: > > datastore2~;iscsiadm list target > Target: iqn.1986-03.com.sun:02:6230a9ff-67d0-e995-9e85-f009c17adab9 > Alias: - > TPGT: 1 > ISID: 4000002a0000 > Connections: 1 > Target: iqn.1986-03.com.sun:01:e00000000000.4b2fd513 > Alias: - > ISID: 4000002a0000 > Connections: 0 > > On server 2, the initiator, I create links with devfsadm and use format to > check the disks. But not only is the disk from server 1 not listed, the disk > I'm exporting on server 2 (c9d1) isn't listed either. > Yes, this is expected. I suspect the complete disk(c9t1) from server 2 was given as a LU. Is that true? If you give only a slice of the disk to COMSTAR, then the disk will show up in format. When a complete disk is given to the COMSTAR framework(using sbdadm create-lu), the COMSTAR framework holds that device exclusively and thus prevents other commands like format from seeing it. > datastore2~;devfsadm -i iscsi > datastore2~;format < /dev/null > Searching for disks...done > > AVAILABLE DISK SELECTIONS: > 0. c8d0 <DEFAULT cyl 60725 alt 2 hd 255 sec 126> > /pci at 0,0/pci-ide at 1f,2/ide at 0/cmdk at 0,0 > 1. c8d1 <DEFAULT cyl 60797 alt 2 hd 255 sec 126> > /pci at 0,0/pci-ide at 1f,2/ide at 0/cmdk at 1,0 > Specify disk (enter its number): > datastore2~; > > Shouldn't I also see the disk I'm exporting locally as well as the disk from > server 1? > > Both machines in the cluster seem to be aware of all the disks: > /usr/cluster/bin/cldevice list -v > DID Device Full Device Path > ---------- ---------------- > d1 datastore2:/dev/rdsk/c7t0d0 > d2 datastore2:/dev/rdsk/c8d0 > d3 datastore2:/dev/rdsk/c8d1 > d4 datastore2:/dev/rdsk/c9d1 > d5 datastore1:/dev/rdsk/c10t0d0 > d6 datastore1:/dev/rdsk/c8t0d0 > d7 datastore1:/dev/rdsk/c8t1d0 > d8 datastore1:/dev/rdsk/c8t2d0 > > Should those disks be visible when running "format"? > The DID paths that cluster shows were created when the cluster node first booted into cluster mode. If you run "cldevice clear", the devices that are no longer accessible will get cleared. thanks, Augustus. -- Augustus Franklin, Solaris Cluster, http://blogs.sun.com/augustus
