If both the primary and replica is missing it is important to figure out the cause so it can be prevented from happening again. Instead of reindexing you could use the cluster reroute API to allocate an empty shard to replace the missing shard (see the allocate_empty_primary parameter), which may be easier.