Issue Description
Some site uses two hosts to make two-node cluster, whose backend is linked to OceanStor S5100, in which 25volumes are mapped to the hosts. Before operating, the system need to be tested by two-node script, in the process two-node switch often occurs, which lead that the file system of the volumes mapped to the hosts is damaged. After operating, when visiting these volumes, the host comes error, ‘Input/Output Error’.
Alarm Information
Some volumes of the host cannot be read and written and comes error, ‘Input/Output Error, the damaged volumes cannot be read and written.
Handling Process
The file system is xfs. You can use xfs repair to fix the error. Before fixing the file system, make sure the file system has been uninstalled, you can use umount to uninstall the prefixed volumes, and use mount to check if the volumes are uninstalled after uninstalling.
The xfs repair format is: xfs_repair [device]. For example, xfs_repair /dev/sdb1 is fixing sdb1, in the process, you will see “can not map block 210”,which means this volume is damaged. Repeat to run xfs_repair until there is no “can not map block 210”.
If there is error in xfs_repair, use xfs_reapir -L [device]
Root Cause
Two-node script test is finished by integrated manufacture, in which process two-node switch cannot be avoided, so it lead to the damage of the volumes. After operating, when visiting these volumes, the host comes Input/Output Error, and the damaged volumes cannot be read and written.
Suggestions
After two-node test, you should recreate file system to avoid damaged volume mounted on the host.
