Hi Listers, I cross-posted this on the LINUX-390 Listserv as well.
We had an issue when microcode was upgraded on all our DS8000s . According to the CE, about 5 hours after the microcode was installed, the two PCHIDs connected to our SCSI DS8000 started posting error messages every 6 seconds. I logged on to one of our test systems and cd'ed to the directory mounted on the scsi disk. When I issued the ls command, my linux went to sleep and never came back. Our fcp is defined using EDEVs...so using MAINT, I then issued the Q EDEV command and it just locked up MAINT. The following messages appeared on OPERATOR (every minute): HCPMHT2150I DASD 0E01 and interrupt is pending HCPMHT6304I IRB= (all zeroes) HCPMHT6305I Userid=LNXCUPT We eventually forced the linux guest down and shutdown z/VM and then toggled the PCHID off/on which cleared up the error. I was able to IPL both z/VM and the linux guest and had no problems getting to the data. Our CE wonders if it was CUIR ... the documentation states it supports z/VM. I opened an issue with IBM to confirm z/VM supports CUIR on SCSI and I've pasted their response below: "We do not support CUIR on SCSI. CUIR only affects CKD volumes. The HW will not give us any quiesce/resume requests for SCSI volumes regardless of whether we have a single or multiple path(s) to the LUN. " Even without CUIR surely EDEVs should handle failures. Responses on the LINUX-390 indicated native SCSI handles path failures just fine. Has anyone experienced an outage to their data on SCSI when defined on EDEVs due to path failures? Thanks in advance for any advice. susan