Trying to configure a dataset to use 4K record size, but getting the above warning. The full warning is as follows:
Previously, I had heeded the warning and configured a 16K record size - despite the fact that I really didn't want a 16k record size. Now I have overridden the warning and configured a 4k record size and am seeing reduced read I/O, and increased write I/O for my workload (both of these are positive as I feel the 16K record size was reading more data than required, and writing larger blocks to disk than required and delaying other I/O writing being written to disk).
My question is, why am I getting this warning and what (if anything) and should I be concerned about reconfiguring something or leaving this as-is. I can't find any information on the web about this warning, and a search of the forums reveals no information or posts.
My topology is as follows: 4x2 disk mirrors (HDD) with a single ZIL/SLOG (SDD). The machine was bought from and configured by iXSystems (FreeNAS Mini XL). I am running this dataset with primarycache=metadata due to the random access nature and large size of dataset (blockchain data > 100GB single file). The dataset is being shared via NFSv3. My intent is to keep this data out of the zfs cache with primarycache=metadata, while maximizing the disks ability to handle more frequent and smaller I/O.
Thank you for your help!
Previously, I had heeded the warning and configured a 16K record size - despite the fact that I really didn't want a 16k record size. Now I have overridden the warning and configured a 4k record size and am seeing reduced read I/O, and increased write I/O for my workload (both of these are positive as I feel the 16K record size was reading more data than required, and writing larger blocks to disk than required and delaying other I/O writing being written to disk).
My question is, why am I getting this warning and what (if anything) and should I be concerned about reconfiguring something or leaving this as-is. I can't find any information on the web about this warning, and a search of the forums reveals no information or posts.
My topology is as follows: 4x2 disk mirrors (HDD) with a single ZIL/SLOG (SDD). The machine was bought from and configured by iXSystems (FreeNAS Mini XL). I am running this dataset with primarycache=metadata due to the random access nature and large size of dataset (blockchain data > 100GB single file). The dataset is being shared via NFSv3. My intent is to keep this data out of the zfs cache with primarycache=metadata, while maximizing the disks ability to handle more frequent and smaller I/O.
Thank you for your help!
