CFSHELF-6299: NSM100/NSM100B/NSM140 shelf disk path loss after PCIe error in NVMe drive leads to reset
Issue
- One or both NSM shelf modules on an NS224 disk shelf unexpectedly reset during normal operation.
- In rare cases, a simultaneous reset of both NSMs can cause ONTAP to lose all paths to the NVMe drives in the shelf, triggering a multi-disk panic (MDP).
- ONTAP EMS logs that rule out this problem:
sfu.downloadSuccess:info– A shelf firmware update will reboot each NSM to install new firmware, the messages below are expected following this event.- ONTAP EMS log symptoms for this problem:
sla.shelf.mod.reboot.unexp:error– Unexpected NSM reboot detected by ONTAP, with a "Startup type 5-Crash" message.nvmeof.timeout:notice– Mass timeout of drives belonging to the same shelf following an unexpected reset.scsi.cmd.selectionTimeout:error– Mass timeout of drives belonging to the same shelf from one or more network adapters.netif.linkDown:info– In a direct-attached storage configuration, one or more links go down as the NSM module resets.
