Volume Pinning Caveats
When performing certain operations related to volume pinning, keep these caveats in mind.
Condition | Solution |
---|---|
Volume Pinning Enablement |
To be able to pin the volume, the cache must be enabled on the performance policy for the volume. |
Quota and Volume Size Changes |
When changing a pinned volume's quota, the amount of cache reserved for the volume will also change. You will not be able to pin volumes until sufficient cache is available. |
Moving a Volume |
When moving a volume, the pinnable capacity on the destination array or pool must be sufficient to pin the moving volume. Before moving the volume, ensure there is enough cache on the destination array or pool. The cache reservation on the source will be freed on the completion of the move. If you attempt to pin a volume during a volume move, pinning will not be guaranteed until a scan has finished on the destination. This could take some time (up to several hours) depending on how much content is in the volume, and how much other activity there is. |
Replicating a Volume |
If you replicate a pinned volume, the volume will not be pinned downstream. |
Adding Shelf Capacity |
When adding a shelf to an array, if the array's new pinnable capacity is less than the size of the pinned volumes, the shelf activation will fail. |
Failed, Removed, or Upgraded SSD |
When an SSD fails or is removed, if the amount of pinnable capacity is less than the current size of the pinned volumes, then all volumes may become unpinned because there is not enough free cache to pin the blocks. A message is displayed recommending that you consider adding capacity or unpinning some volumes to restore performance to cache pinned volumes. If the SSD is replaced with one of the same capacity, volume pinning continues normally. If the SSD is replaced with one of a different capacity, the amount of pinnable cache is recalculated. One consolidated alert is sent for all volumes when the free usable cache drops below the acceptable level, and another alert is sent once the free usable cache returns to an acceptable level for the rescan to be completed. |
Pinning an Existing Volume |
If you are using the Edit function to specify that a volume is to be pinned, this initiates a scan of the volume. Pinning will not be guaranteed until the scan of the volume is finished. This could take some time (up to several hours) depending on how much content is in the volume, and how much other activity there is. Alerts are sent both when pinning begins and ends, on a per-volume basis. |
Performing a Bin Migration or Pool Merge |
If you have performed a bin migration (for space balancing, for example), or if you attempt to pin a volume during a bin migration or pool merge, pinning will not be guaranteed until a rescan has been completed on the bin's new destination. This could take some time (up to several hours) depending on how much content is in the volume, and how much other activity there is. For a bin migration, one consolidated alert is sent for all volumes once the rescan is finished. For a pool merge, one consolidated alert is sent for all the volumes before and after the merge. |
Performing a Volume Snapshot Restore |
If you want to do a volume snapshot restore (unpin an old volume and pin a new volume) to re-establish the heat map (cache hit information), pinning will not be guaranteed until a scan of the new tip has finished. This could take some time (up to several hours) depending on how much content is in the volume, and how much other activity there is. Note: A snapshot restore can only be performed after
unpinning the volume.
|
Performing a Software Upgrade |
In case of a software upgrade, pinning will not be guaranteed until a rescan is performed to determine whether any blocks that were in memory were evicted as a result of the software upgrade. This could take some time (up to several hours) depending on how much content is in the volume, and how much other activity there is. One consolidated alert is sent for all the volumes regarding possible loss of pinning, and another alert is sent once the rescan has completed. |
Unmanaged Shutdown |
In the case of an unmanaged shutdown (where a few segments of pinned data are not flushed and can show up as cache misses), pinning will not be guaranteed until a rescan of the system is finished. One consolidated alert is sent for all the volumes before and after the rescan. |
Volume Promotion/Demotion |
Demoting a pinned volume will automatically unpin the volume. The volume can be pinned again after promoting it. |