HPE 3PAR Thin Provisioning: Space Commands That Matter

HPE 3PAR and Primera arrays are thin by default, which means the array reports several different space numbers at once and none of them means what a provisioning spreadsheet assumes. This article covers the handful of CLI commands that answer the only three questions that matter: how much is actually written, how is each CPG laid out, and is any volume growing towards a limit.

Reserved versus used

cli% showvv -s
cli% showvv -s VV1
cli% showvv -v VV1

showvv -s returns per-volume space accounting. The columns to read are Usr_Rsvd_MB (space reserved for the volume's user space), Usr_Used_MB (blocks actually written) and Usr_Free_MB (remaining headroom in the reservation). A thin volume whose Usr_Used_MB tracks its Usr_Rsvd_MB is close to needing attention; a volume where reserved is far larger than used is sized for a peak it has never reached. showvv -v adds the policy detail — provisioning type, allocation policy and the snapshot space settings.

CPG layout

cli% showcpg -sdg
cli% showcpg -space CPG_NAME
cli% showspace

Common Provisioning Groups are where thin volumes actually draw their space. showcpg -sdg breaks each CPG down by device type and RAID level, which is how you discover that a volume intended for SSDs is silently growing on nearline disks because both device types were included in its CPG. showspace gives the array-wide free-space picture to compare against.

Where the space went

cli% showvv -perc
cli% showvv -cpgalloc -s
cli% showvvpd VV1
cli% showvvmap -v VV1

showvv -perc shows the space breakdown by category for each volume, which is the fastest way to find a volume whose snapshot space has quietly outgrown its user space. showvvpd shows the physical disk distribution if you need to confirm the data really landed where the CPG design intended.

Acting on what you find

cli% growvv VV1 100g          # extend user space
cli% tunevv usr_cpg NEWCPG VV1 # move user space to a different CPG
cli% compactcpg CPG_NAME       # reclaim unreferenced space in a CPG

Two operational rules keep thin provisioning honest. First, monitor showvv -s for volumes whose used space approaches their reserved space and set alert criteria before, not after, the first write failure. Second, remember that deleting data inside a guest does not return space to the array — the CPG must be compacted (or the volume tuned) before the freed blocks become usable again.

Quick health check order

1. checkhealth -svc for overall state.
2. showalert -n for anything outstanding.
3. showpd -failed -degraded for disk problems.
4. showcpg -sdg and showspace for layout and headroom.
5. showvv -s for per-volume growth.

Related reading: HPE 3PAR CLI cheat sheet, 3PAR createvv, createhost and present vlun, PowerStore PSTCLI volume provisioning.

原文链接：https://www.spectra.com/hpe-3par-common-cli-command-line-interface-commands/