Issues Fixed in Lightbits 3.4.2

With el9 5.14 kernels, multipath nvme_core parameter is enabled by default.

How to check:

$ grep -rH '' /sys/module/nvme_core/parameters/multipath

/sys/module/nvme_core/parameters/multipath:Y

If an SSD to be used by Lightbits is multipath-enabled (has a non-zero identify-ctrl field CMIC),

the SSD will not be recognized and GFTL will fail to power up.

How to check:

$ for i in /dev/nvme_n_; do echo $i $(nvme id-ctrl $i|grep cmic); done

/dev/nvme0n1 cmic : 0x3

/dev/nvme1n1 cmic : 0

/dev/nvme2n1 cmic : 0

/dev/nvme3n1 cmic : 0

/dev/nvme4n1 cmic : 0

/dev/nvme5n1 cmic : 0

In order to fix this, disable multipath; e.g.:

$ echo "options nvme_core multipath=N" >> /etc/modprobe.d/nvme.conf

and remove and reload the module or reboot the server for the setting to take effect.

VCP traffic monitoring graphs are incompatible with newer Lightbits cluster releases.

Traffic metrics will not be available, or will be incorrect with VCP 1.3 (or older) connected to a Lightbits cluster running releases 3.4.1 or newer.

Specifically, the following graphs will have missing/incorrect information with the VCP/Lightbits cluster combinations mentioned above:

Under the cluster/Monitoring tab:

• Cluster IOPS will not be available.
• Cluster bandwidth will not be available.
• Volume IOPS will not be available.
• Volume bandwidth will not be available.
• Cluster latency will be incorrect (the displayed value needs to be multiplied by CPU frequency to get the correct value).
• Volume latency will be incorrect (the displayed value needs to be multiplied by CPU frequency to get the correct value).

Discovery using nvme-cli (nvme discover, nvme connect-all) will not work with nvme-cli versions 2.0 and above.

With nvme-cli below 2.3, this will result in nvme segfault.

With nvme-cli 2.3 and above, it will fail with the following error: Failed to get discovery log: Invalid slot.

To work around this issue, use Lightbits' discovery-client or downgrade your nvme-cli version.