feldsam wrote:
Hello, from my experience, when I was playing with scsi discards, I triggered reclaim by running a compact_cpg task. This is not running by default, so I scheduled them.
In that case you got it all wrong.
Compact cpg releases free capacity from the CPG to the system.
Reclaim releases free capacity from the volumes to the CPGs.
Reclaim is running as a hidden task triggered every five minutes on a limited amount of volumes (I think 8)each run. If reclaim "fails" (ie it can't reclaim anything from a volume) and the volume has more than 4GB of free capacity per node in the system it will trigger a defrag task on the volume (visible).
Compact CPG is usually a manual task. The only way I know it autotriggers is if you're doing AO and hits certain parameters. In most cases compact cpg isn't really needed. Remember that the CPG will use the free capacity in the CPG before allocating new capacity. So in most cases where you are expecting the CPG to grow at some point and you're not low on capacity with multiple CPGs on the same tier, compact cpg is simplity wasted IOs.