On 04/02/2010 21:26, Ellard Roush wrote: > Hi Robert, > > It is good that you already knew about the "ZFS and Quorum" issue. > > Well driver/firmware/device issues are not that uncommon. > Some time back my frustration with these problems > motivated a change. Sun Cluster software now performs a > read after every quorum related write to a SCSI-2 device to verify that > the information actually is on the SCSI-2 device. Sun Cluster also > logs every quorum related write operation. So you now have > proof of what was written to the quorum device and when that happened. > This gives you the evidence that you will need to convince > the driver/firmware/device people that the bug is in their area. >
yeah, great. unfortunatelly the buffer doesn't hold the information I need anymore. anyway, I will investigate further. > Your latest email points to another possible problem. > You mention that multiple clusters connected to the same > storage array. Only one cluster can manage any particular > disk. The fencing or quorum actions by another cluster will > cause problems. So make sure that both clusters are not > touching the same disks. If multiple clusters are accessing > the same storage array, there is a good chance that this > problem is happening. The admin can issue a command to tell > a cluster not to fence a particular device. The admin also > controls which device is the quorum device. Thus the admin > can ensure that a device is NOT a quorum device for > multiple clusters. > of course, LUN making along with SAN zoning is used to assure that it doesn't happen.
