Replies: 1 comment 2 replies
-
No need for this, use chunkserver labels and classes for it like we used to do ;) https://moosefs.com/Content/Downloads/moosefs-storage-classes-manual.pdf |
Beta Was this translation helpful? Give feedback.
2 replies
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
-
hi,
Chunkserver uses mfshdd.cfg to configure where it store its data.
How about adding new marking there, marking disk as 'cache' - in understanding that when data in uploaded to chunkserver it is saved on 'cache' disk, and after successful upload it can be redistributed to other disks.
We could then have server with X HDD disks, and Y SSD. SSD could be marked as 'cache', so upload will be fast (as data will be stored in SSD) and then after data is 'safe' it will be distributed to 'permanent/slower' disks.
Currently such solution requires dedicated chunkserver with only SSD drives and using 'CREATE' storage class, and after uploading file, it is using network to redistribute data.
Beta Was this translation helpful? Give feedback.
All reactions