Unkillable processes in state D when all paths to a volume are down

This document (7001880) is provided subject to the disclaimer at the end of this document.

Environment

SUSE Linux Enterprise Server 10
SUSE Linux Enterprise Server 9

Situation

Processes waiting for I/O (i.e., processes in state D) don't get killed with 'kill SIGKILL' when all the paths to the LUN are down, to which the I/O process is writing to. The device mapper is configured to use the 'queue_if_no_path' option for multipathed volumes.

Resolution

The 'queue_if_no_path' device mapper setting causes the kernel to wait indefinitely for I/O operations to complete, even when all pathes to a volume are gone. Subsequently, all processes accessing the LUN are not killable and remain in state 'D' until the volume returns.

The device mapper can be set to return an I/O error in case a volume becomes unreachable with the command
'dmsetup message DEVICE_NAME 0 fail_if_no_path' (replace DEVICE_NAME with the device name of the volume.)

Disclaimer

This Support Knowledgebase provides a valuable tool for SUSE customers and parties interested in our products and solutions to acquire information, ideas and learn from one another. Materials are provided for informational, personal or non-commercial use within your organization and are presented "AS IS" WITHOUT WARRANTY OF ANY KIND.