Will FAST/VP fill up a thin pool?

Please consider moving this question as-is (no need to recreate) to the proper forum for maximum visibility.  Questions written to the users' own "Discussions" space don't get the same amount of attention and questions can go unanswered for a long time.

You can do so by selecting "Move" under ACTIONS along the upper-right.  Then search for and select: "Symmetrix Support Forum" which would be the most relevant for this question.

FAST VP stops promoting tracks to the pool when its 90% full. This is a default behavior and can be changed by changing the PRC values of the pool.

regards,

Saurabh

FAST VP will not fill up a pool completely due to the PRC setting on each pool.  The PRC (pool reserved capacity) reserves a certain percentage of each pool that cannot be used for FAST VP movements.  By default, the PRC is set system-wide with a value of 10%.  What this means is that when a pool is 90% full, FAST VP will no longer move data into that pool. It can, however, continue to move data out of the pool.

The system-wide PRC setting can also be overidden by setting a PRC on each individual pool.  The best practice recommendation is to set the PRC to 1% on pools that have no devices bound to them.  On the pools where devices are bound, the PRC should be set higher to reserve capacity for new allocations resulting from host writes.

Keep in mind, also, that 5876 has a feature called VP allocation by FAST policy.  This allows new allocations from host writes to come from any pool in the policy the device is associated with.  So, even if the FC pool fills up, a new allocation could come from either EFD or SATA.  As long as the array is not oversubscribed (the bound pool can be) all new writes will be successful.

SYMCLIGuy wrote: The system-wide PRC setting can also be overidden by setting a PRC on each individual pool.  The best practice recommendation is to set the PRC to 1% on pools that have no devices bound to them.  On the pools where devices are bound, the PRC should be set higher to reserve capacity for new allocations resulting from host writes.

does/how this recommendation change if i am using thin, fully allocated devices that are all bound to FC pool (in 3 tier configuration). Should i lower PRC for EFD and SATA to 1%, current is at default 10%.

Thanks guys this is good stuff, I will lower the EFD pool PRC to 1% and I also turned on allocate by policy.  Thanks for the clarification and advise.

yeah, it's one of those discussions that always comes up, benefits of oversubscribing and disastrous consequences of running out of space and not being able to procure capacity fast enough.

I do like recommendation of lowering PRC, especially for my EFD pool.  Was this documented in the latest FAST-VP paper ?

The 1% PRC for pools with no devices bound has been the recommendation has been in the FAST VP best practices since the first version :-)

Regarding your comment on preallocating to prevent not running out of space - I assume this means that you are subscribing the array 100% (you can't preallocate what you don't have). If that is the case, you should make sure you enable allocation by policy in 5876 code. If the capacity of TDEVs bound is not greater than the capacity of the array then you won't run out of space even if you don't preallocate.

i am on 75 and not going to 76 anytime soon (very conservative customer)