Symptoms
The host is directly connected to disk arrays with an optical fiber. After the following operations, I/Os suspend periodically but do not stop:
Run the dd command to read a LUN of controller A.
Switch the optical fiber from port 2 to port 0 of controller A. At this moment, I/Os failed over to the path of controller B.
Run the cfgmgr command. MPIO sets the path connecting port 2 of controller A to Failed and generates a new path for port 0 of controller A. Then I/Os fail over to controller A.
Analysis
Fault location excludes the possibility of disk array or UltraPath (developed based on the MPIO framework) fault, so it is believed that the problem is caused by the built-in defect or mechanism of MPIO.
Solution
The root cause cannot be located as MPIO is completely closed, but you can avoid this problem by deleting the invalid path.
Run lspath -F "path_id:connection:parent:status" -l hdisk2 to list path information, as shown in the following figure.
As shown in the output, colons (:) are used to separate columns. The first column indicates the path ID. The characters in the second column are further divided into two fragmented parts by a comma (,), indicating the connection. The third column indicates the parent. The fourth column indicates the path status. The figure shows that the path connecting to port 2 of controller A (0:2201001882552bcf, 0:fscsi0:Failed) is invalid.
2. Run rmpath -l hdisk2 -p fscsi0 -w 2201001882552bcf,0 -d to delete the invalid path, as shown in the following figure.
