After the content distribution to a local disk, events from the OSD
op_commit: op submitted by the main OSD (ie written to the log)
op_applied: OP has been written () 'en to FS backup on the primary server (that is, applied to the memory but not flushed to disk)
sub_op_applied: op_applied, but for a copy of "subop"
sub_op_committed: op_commited, but subop for copies (only for EC pool)
sub_op_commit_rec / sub_op_apply_rec from: when it hears the message above, the main mark of this, but for a particular copy
commit_sent: we will reply back to the client (or main OSD, for sub-operations)
of these events appears to be a lot of redundant, but an important border crossing internal code ( such as cross-lock to pass data to the new thread).
