Can someone explain how the recovery steps of elasticsearch works?
e.g. when a node comes back up, shards get initialized. I see in cat recovery files files_percent & translog and translog_percent
translog recovery makes sense to me [reapply everything in translog], but what is file_percent? I know each shard is made of many segment files. But what exactly is initialization step of a shard? What operations are performed during initialization? And why is it slow at times?