HPE Storage Users Group

I've been trying to wrap my head around this one for a bit. I did some clean up this weekend, zeroing out filesystems on my Linux VMs and I noticed that I was able to lower my Used User Size space but I didn't dent my Reserved User Size at all.

Does anyone know what exactly (and how) the Reserved User Size is calculated?
Also, what is the actual impact of this reserve size? For example if I have a 2 TB VV and I have 608GB Used User Size and 901GB Reserved User Size, does the sum of those two values show me what I am actually using?

Hello Rydley,

So reserved user size is how much is "allocated" to the VV. Allocated space contains "User Used" and "User free" space. VVs will grow in increments determined by the Inserve version, # of nodes, and raid setting.

Have you tried compacting the CPG after the zero_detect process was done at the VV level?

The reclamation process will not grab back every available chunk let of capacity from a VV... (I am not sure what the exact number is... but I am going to guess 10%), 10% or so of free space is ignored by the thin reclamation process. That way VV growth has a place to go right now, instead of having to (re)allocate room, in realtime, for new writes.

Rydley wrote:For example if I have a 2 TB VV and I have 608GB Used User Size and 901GB Reserved User Size, does the sum of those two values show me what I am actually using?

To directly address your example:

608 is used of the 901 that is reserved, leaving 293 free for new growth before a new allocation/reservation occurs.

Rich,

So my Allocated Capacity on the CPG these TPVVs sit on will not change unless I grow beyond the reservation?

Will the Reserved User Size ever shrink? I imagine this will not be a real-time process and may occur overtime as to not impact performance, if it does shrink at all.

Thanks for the prompt response,
Rydley

Correct, your CPG should not grow until one of the VVs inside it needs more capacity reserved. Reserved User Size can shrink, you just have to free up enough "contiguous" space. Thin Reclamation will covert VV Used to VV Free in 16kb increments, but will only convert VV free to CPG Free in contiguous 128m increments. Perhaps a defrag prior to zeroing may yield better results?


According to the 2.31 concepts guide:

TPVVs are capable of responding to host write requests by allocating space on demand in small increments, beginning with 256 MB per controller node supporting the TPVV. These allocations are adaptive since subsequent allocations are based on the rate of consumption for previously allocated space. With 256 MB per node as the default, allocations increase by increments of 256 MB per node as the system demands. For example, if a TPVV is initially allocated 256 MB per node but then consumes that space in less than sixty seconds, the next allocation becomes 512 MB per node. However, if the initial 256 MB per node is consumed more slowly, the next allocation increment remains at 256 MB per node. Under this provisioning scheme, the maximum allocation increment is 1 GB per controller node supporting the TPVV.

...

Freed blocks of 16 KB of contiguous space are returned to the source volume, freed blocks of 128 MB of contiguous space are returned to the CPG for use by other volumes.

Interesting. I am seeing a huge discrepancy in the range of 400GBs between what the VMs Guest OS filesystems are using (tallied up manually) and what 3PAR is reporting as User Used.

/wrists