- hadoop: clean up assert usage
mds
+- special case commit of stray dir to avoid having to commit, re-commit strays?
+ - once we commit one stray, we have to re-commit later to remove it, which means we commit other new ones.
- put inode dirty fields into dirty_bits_t to reduce per-inode memory footprint
- don't sync log on every clientreplay request?
- pass issued, wanted into eval(lock) when eval() already has it? (and otherwise optimize eval paths..)