is_zk_lock_held incorrectly returns false on error
derekjn opened this issue · comments
If the zoo_exists call fails in is_zk_lock_held as the result of a transient error, is_zk_lock_held will incorrectly return false, causing the caller to potentially wait on itself for a new lock. We should retry a few times instead, and if a definitive lock check can't be completed, the process should error out.