A host may lose access to volumes after PowerStore upgrade to the affected version.
Alternatively, if PowerStore was previously upgraded to the affected version without any issues, updating the host drivers to a version supporting NVMe exposes this issue.
After the hosts are updated to support NVMe, a second process login (PRLI) can occur from the host following each link up event. This second PRLI is for NVMe transport, and the PowerStore does not handle this second PRLI optimally.
This issue is addressed in PowerStoreOS v3.
Workaround:
Disable NVMe support in the driver.
- For ESXi hosts, see VMware article 84325
. Summary of VMware article is below:
ESXi using QLogic or Marvel HBA and qlnativefc driver:
esxcfg-module -s 'ql2xnvmesupport=0' qlnativefc
ESXi using Emulex HBA and lpfc driver:
esxcli system module parameters set -m lpfc -p lpfc_enable_fc4_type=1
- For Red Hat hosts, see Red Hat article 3522911
. Summary of Red Hat article is below:
Linux hosts using QLogic or Marvel HBA and qlnativefc driver, add or edit the line below:
vi /etc/modprobe.d/qla2xxx.conf
ql2xnvmeenable=0
Linux hosts using Emulex HBA and lpfc driver, edit the following file and add or edit the line below:
vi /etc/modprobe.d/lpfc.conf
lpfc lpfc_enable_fc4_type=1
NOTE: The hosts must be rebooted after changing the HBA driver configuration.