Troubleshooting
Throughput is lower than expected
Both checks below read the Config section of get-smart:
$ xcena_cli im get-smart 0
...
Config:
ssd_link_speed : Gen5
ssd_link_width : x4
ssd_io_queue : sufficient
The PCIe link to the SSD
Confirm the link came up at the speed and width the SSD is rated for. Anything below what you expect means the link did not train as expected — the SSD cannot reach its rated throughput no matter how many I/O queues it offers. Re-seat the SSD and the MCIO cable, and check the connection against Getting Started.
The SSD’s NVMe I/O queues
InfiniteMemory drives the SSD across many NVMe I/O queues. An SSD that offers fewer queues than the device asks for is still brought up, but with fewer of the device’s storage engines in service — so sequential and random throughput both fall short of the datasheet figures. ssd_io_queue carries the verdict:
| Value | Meaning |
|---|---|
sufficient | The SSD can drive every storage engine — full performance. |
insufficient | The SSD offers too few I/O queues; the device runs a reduced set of storage engines. Use an eSSD that supports more than 16 I/O queues for full performance. |
If the problem persists
When none of the above accounts for what you are seeing, collect the diagnostic logs and send them to the support team along with a description of the symptom. The SDK’s Collecting Troubleshooting Logs covers the troubleshooting.sh script, which gathers the system and device logs in one pass.