We had a session with EMC to have a look at the thin/thick provisioning issues. Apparently, this is by design.
I ended up re-creating all the LUN's which wasn't really necessary but I wanted to know how it works.
When you create a LUN, storage is claimed from the pool and made available for the SP on which you created the LUN. There is however some extra space claimed in the back end.
So in our case:
For storage pool Performance_RAID5 with a total capacity of 7.998TB
Total: 7.998TB
7.998 / 2 = 3.999 /2 = 1.999TB (We divide by two to balance the load on the SP's.)
- 1.999TB (iSCSISPA_RAID5_01)
- 1.999TB (iSCSISPB_RAID5_01)
7.998 - 1.999 - 1.999 should be 4000. However, SPA and SPB claimed some extra space from our pool by creating the first LUN's. There was only 3.880TB left to provision. So 3.880 / 2 = 1.940TB
- 1.940TB (iSCSISPA_RAID5_02)
3.880 - 1.940 should be 1.940 but again SPA claimed some extra space from the pool. Now, there was only 1.827TB available for SPB.
- 1.827TB (iSCSISPB_RAID5_02)
We lost 120GB by creating iSCSISPA_RAID5_01 and iSCSISPB_RAID5_01. So each SP claimed some extra 60GB of storage from the pool.
Creating the second LUN on SPA made that there was only 1.827TB left for SPB. We lost another 113GB.
In the end, we lose +/- 225GB storage on our RAID5 pool.
Or is it 7.998 - 1.999 - 1.999 - 1.940 - 1.827 = 233?
I really hope somebody can further explain. Or correct me if I'm wrong.
Grtz
Filip