[Environment Configuration]
Dorado 6000 V6 NVMe(512GB, 8 * 3.84 TB NVMe), one RH2288H V3 CentOS 8.2 host, which is directly connected to the storage device through 10GE ports.
[Problem Description]
The host multipathing software is set to the round-robin 0 mode, and the reliability test results show that the host multipathing software is restored to 0 within 4s.
1, IOPS 20W, restart the controller from developer mode: IO stopped 4s
2, IOPS 20W, restart the controller from minisystem: IO stopped 4s
[Optimization Solution]
1. Change the default multipathing mode to service-time and restart multipathd.service.
systemctl restart multipathd.service systemctl enable multipathd.service
2. Change the default iSCSI timeout interval to 120s. Run the following command to change the timeout interval to 120s. (The value of new_value must be set based on the service requirements, for example, 10, in seconds.)
iscsiadm -m node -o update -n node.session.timeo.replacement_timeout -v new_value
Test result after modification:
Test result (service time 0):
1, IOPS 2W, remove the controller: no IO stop
2, IOPS 20W, remove the controller: IO stopped 1s
3, IOPS 20W, restart the controller from developer mode: no IO stop
4, IOPS 20W, restart the controller from minisystem: no IO stop
5, IOPS 20W, remove the SmartIO card: no IO stop
