Hi
I need some confirmation/details on the UNMAP/reclaim feature in 5.5
- While running unmap we see high Read rates and low/zero write rates on the storage adapters. One would expect high write rates because the unmap command writes zeroes to storage?
- After unmap is finish it seems the host runs a rescan of vmfs volumes. Is this by design of the unmap command? I found this in the hostd.log very soon after completion of unmap:
2015-10-21T09:46:37.628Z [26BC1B70 verbose 'Hostsvc.FSVolumeProvider'] RefreshVMFSVolumes called
2015-10-21T09:46:37.629Z [26BC1B70 verbose 'Hostsvc.FSVolumeProvider'] RescanVmfs called
We suspect that rescan causes a glitch in performance graphs as shown here (snip of storage adapters, happens also on the other graphs) :
As the graph shows it doesn't happen after every unmap run and I have experimented with different block values when running the unmap command, but I can't find any consistency in it...
The commands have been run on a esxi host (5.5 up2) running on HP BL460 Gen9 against HP 3Par storage (VMFS 5) through FCoE
Any clarification on this is much appreciated!
Regards,
Rudi