Recovering from a destroyed node

[Jason_Baumgartner]

I had a node that was destroyed that contained shards from various indexes. I have backups of the original data and need to reindex the data that was held in those shards. I have wiped the shards affected so that they are now empty and on a new node.

My question is this -- if there is an empty shard with other shards that contain data in an index, if I reload data into the index, will the empty shard fill up first (to balance with the other shards?) If not, how do I go about making sure the empty shards load with data so that they are balanced with the existing shards that do contain data. There is no routing set up for these shards / indexes.

Thank you!

[warkolm (Mark Walkom)]

It's not guaranteed without routing, because it will use the hashing algorithm to spread things out and I am not 100% sure that it'll deterministically put things back where they were.

[Christian_Dahlqvist]

If the index has some shards with data and you reindex the full data set, you will get duplicates unless you are specifying a custom id for the documents. If you use auto generated ids the new data will spread quite evenly across the available shards.