This is an observation with hammer that I've noticed for a long time and it
happens often enough that I want to put it out there and see if there's
anything to it. I have a number of master/slave pairs and I typically set
them all to:

snapshots 1d 7d
prune     1d 5m
rebalance 1d 5m
#dedup   1d 5m
reblock   1d 5m
recopy    30d 10m

I run the standard nightly hammer maintenance script. Typically, shortly
after the 7d period, the master volume will recover a big chunk of space,
but the slave won't. That's fair, I'm aware of pruning and reblocking
behavior. However, the slave will more often than not take a week or more
after that to recover that same space. Is there something about the way
mirroring writes blocks to disk that would have this effect?

Tim

Reply via email to