Reserved User size more than 100%

[scoggins]

I don't expect it to give back all the space. But when you have a 2Tb VV that only has 800gig. I'd like some of my 1TB free back!

Code: Select all

3PAR cli% showvv -s *vdi*
                             ---Adm--- ---------Snp---------- ------------Usr------------                 
                             --(MB)--- --(MB)--- -(% VSize)-- -----(MB)----- -(% VSize)-- ------(MB)------
 Id Name           Prov Type Rsvd Used Rsvd Used Used Wrn Lim    Rsvd   Used Used Wrn Lim Tot_Rsvd   VSize
206 prd.esx.vdi.01 tpvv base 1536 1259 1024    0  0.0   0   0 1946240 871272 41.5   0   0  1948800 2097152
----------------------------------------------------------------------------------------------------------
  1 total                    1536 1259 1024    0              1946240 871272               1948800 2097152

As you can see reclaim has worked. It knows the space is free. But it never drops the reserved space. So I have 1.1Tb tied in this VV that I can't use anywhere else...

The "solution" that engineering gave me was to do a physical copy of the VV and drop the old pun, represent the new one (aka downtime of the VV). Doing a physical copy of the VV only copies the used space so you end up with a properly thin copy.

[Cleanur]

As per my previous post use a thin to thick to thin conversion, then back to thin. That way you don't need to clone and re-present anything, so no downtime. Of course if you don't have DO (not sure why people don't take DO, it's just so useful) then maybe you can ask your account team for a temp license.

[scoggins]

As per my previous post use a thin to thick to thin conversion, then back to thin. That way you don't need to clone and re-present anything, so no downtime. Of course if you don't have DO (not sure why people don't take DO, it's just so useful) then maybe you can ask your account team for a temp license.

*nod* I've started doing this on some of our VVs already.

[Richard Siemers]

Ahh I see where you're at now.... I thought the next step is to compact the CPG that contains the VV?

Disclaimer: I have only JUST updated my Inserv's to 3.1.2 and have not performed this process since 2.3.1.

[apol]

We have the same problem, and 3TB difference between used and reserved space per vv is a little too much to just have some space for writes coming in...

How we got there:
Every 3-4 months, ESX-admins and we storage guys compare the volumes-statistics. Usually, 3par has smaller usage, approx. 30-50% of what the esx-guys see.

But everytime, there are 3 or 4 volumes where 3par sees more data than esx sees --> those vvs get treated with sdelete.

The problem, as written above: We have some vvs where the extra space did not get reclaimed in the 3 months since we sdeleted. Very small amounts of space got reclaimed on the primaries every once and a while, but it stopped. And on the secondaries (sync rc), the difference betweend used and reserved space is even bigger.

I'm glad this is a known bug fixed in 3.1.3. We are awaiting our update to 3.1.3 MU1 since the update to 3.1.3 was cancelled because of known issues with remote copy...

[RitonLaBevue]

I the volume grows beyhond 100% when using sdelete then you maybe don't use the right switch (c / z)
With Windows 2008 you could use this tool, faster than sdelete:
http://www.antemeta.com/en/Reclaim-Tool

Care to not use if if you already got high latencies...

[Isaac R]

I don't see any improvement with 3.1.3 MU1 and this is quite disappointing.