I was reading this http://www.vmware.com/pdf/vi3_san_design_deploy.pdf
on how ESX manages paths. On page 51, 61 in the pdf, it says that "esx performs a path failover if a SCSI I/O request fails with a FC driver status of NO_CONNECT." If I hook up a Xgig Analyzer to my storage array and ESX server, will I see NO_CONNECT in the trace? Has anyone tried?
Well it talks about "FC driver status[/b]" (status, not command) - I guess it means error handling code within the device driver, not the FC traffic.
If a path is down, e.g. due to a disconnected cable, the array just cannot[/b] respond back to the adapter anyway.
The other possible situation is when a failover happens within an active/passive storage array. In that case I/Os to the former active path of a SCSI LUN will be answered with "NOT READY" status messages by the storage array, as far as I can tell.
An AlphaServer's console subsystem shows standby(passive) paths as "not connected".
This was covered at TSX last week in the Troubleshooting section the Speaker Cormac went over LUN failover as well as Path Thrashing etc..
He confirmed your question.. The Failover occurs if ESX see the SCSI No_Connect Command..
I did a trace and never found a NO_CONNECT cmd.
In the SCSI command spec I don't see a "NO_CONNECT command" either
Well it talks about "FC driver status[/b]" (status, not command) - I guess it means error handling code within the device driver, not the FC traffic.
If a path is down, e.g. due to a disconnected cable, the array just cannot[/b] respond back to the adapter anyway.
The other possible situation is when a failover happens within an active/passive storage array. In that case I/Os to the former active path of a SCSI LUN will be answered with "NOT READY" status messages by the storage array, as far as I can tell.
An AlphaServer's console subsystem shows standby(passive) paths as "not connected".