Ideas for mount and mkfs options

This:
* ensures no lazy creation post mkfs
* removes the journal, apparently this speeds up writes a little
* set FS block size to 4k, to match NVMe block size, should stop
  some fragmentation that could happen a lead to slow down that is
  fixed by trim
* explicit discard during fs creation (apparently its the default)
* explicit discard on mount, keep freeing blocks as we delete things
  so we don't need to re-create the FS so often to get performance back

Author: JohnGarbutt

fs-ansible/roles/beegfs/tasks/format-mount.yaml

@@ -16,7 +16,7 @@
   tags: ['never', 'format', 'mount', 'unmount', 'stop_all']
 
 - name: Format disks
-  command: "mkfs.ext4 -i 8192 -I 512 -J size=400 -Odir_index,filetype /dev/{{ item }}"
+  command: "mkfs.ext4 -E lazy_itable_init=0,lazy_journal_init=0,discard -b 4096 -i 8192 -I 512 -Odir_index,filetype,^has_journal /dev/{{ item }}"
   loop: "{{ all_disks }}"
   tags: ['never', 'format']
 
@@ -31,7 +31,7 @@
       loop: "{{ all_disks }}"
 
     - name: Mount EXT4 OSTs
-      command: mount -onoatime,nodiratime,nobarrier,user_xattr /dev/{{ item }} /data/{{ fs_name }}/{{ item }}
+      command: mount -onoatime,nodiratime,nobarrier,user_xattr,discard /dev/{{ item }} /data/{{ fs_name }}/{{ item }}
       register: command_result
       failed_when: "command_result.rc != 0 and ('is already mounted' not in command_result.stderr)"
       changed_when: "command_result.rc == 0"