I created lvconvert --type cache and added a lv01 (/dev/sda) after that I cannot add another disk to the same cache… (i’m not using cache pool).
PS: I can do it with bcache easy…
perhaps https://www.nocser.net/clients/index.php/knowledgebase/474/Adding-SSD-Cache-to-Existing-LVM.html
Whats the best setup only to cache the metadata of the ext4
Would it make sense and is there one?
I have made exp. with writecache devices (NVMe samsung ent/pro)
they die on 50%TBW a lot
The only known way is to use a LVM, as mentioned above.
Or use ZFS.
I will wait for @elek and his badger cache and execute fsck -fD /dev/sdX1 once offline
and I relied to his thread how to solve the directory fragmentation, hope he reads it.
The main issue is not the file walker or the filesystem itself, but that the filesystem is used as a hash-set. FS that are using btree’s will perform better, but they still apply some kind of optimizations to be able to list all files inside a folder and when the files are not added sequentially this will near always fragment it.