Configuration: Device Mapper Multipath for Linux 187

Limitations and Known Issues

In certain error conditions with the no_path_retry or the
queue_if_no_path feature is set, applications may hang. To overcome
these conditions the you must enter the following command for each
affected multipath device:
dmsetup message [device] 0 "fail_if_no_path"
where
[device]
is the multipath device name (for example. mpath2; do
not specify the path)
I/O may hang when a Device Mapper device is deleted before the volume
is unmounted.
If the scsi_dh_rdac module is not included in initrd, slower device
discovery may be seen and the syslog may become populated with buffer
I/O error messages.
I/O may hang if the host server or storage array is rebooted while I/O is
active. All I/O to the storage array must be stopped before shutting down
or rebooting the host server or storage array.
After a failed path is restored on an MD3600i Series array, failback does not
automatically occur because the driver cannot auto-detect devices without
a forced rescan. Run the command
rescan_dm_devs
to force a rescan
of the host server. This restores the failed paths enabling failback to occur.
Failback can be slow when the host system is experiencing heavy I/O. The
problem is exacerbated if the host server is also experiencing very high
processor utilization.
The Device Mapper Multipath service can be slow when the host system is
experiencing heavy I/O. The problem is exacerbated if the host server also
experiences very high processor utilization.
If the root disk is not blacklisted in the multipath.conf file, a multipathing
node may be created for the root disk. The command
multipath –ll
lists vendor/product ID which can help identify this issue.
book.book Page 187 Tuesday, September 27, 2011 3:25 PM