I have a host with a single cache LV (~800GB SSD in front of a multi-TB RAID array; write-through mode). I ran a routine scrub on the RAID last night (lvchange --syncaction check
). Ever since that completed, the system has been running at 100% I/O capacity; iostat
tells me it's reading data from the SSD and writing to the RAID.
The Cpy%Sync
column of lvs
output for the cache LV (not the RAID) started out at 98% or so before the scrub, which was worrisome in itself, but now it is steadily going down, having currently reached about 70%.
I thought Cpy%Sync
for a cache LV indicated the amount of data on the cache that also existed on the backing store, so in write-through mode it should always be 100%. What does it actually mean and what is the system doing that's making it go down steadily?
There is nothing helpful in dmesg.