Lots of page writeback io errors can tie up a system with syslog writing mostly dm-mpath messages to its log files if KERN_WARNING is directed to do so. In my use case, I had a bunch of Symmetrix LUs with read-only access being written to via dd to the block device file system. The resulting write behind errors were using up lots of cpu time writing on average 50 dm-mpath syslog warning messages for each scsi disk io error message logged. This multiplier reflected the average bio coalescence rate during this period for the 4k bios. Not clear if this is a concern or not ... just thought I would mention it since syslog burned us before.
