I am performing a reindex operation from one index to another within the same cluster running ES 7.5.1. I am using slices=32 and I thought this meant that 32 scroll contexts would be created and used in parallel to pull data from the source to the destination.
However, when running the reindex, I am seeing hundreds of scroll contexts being created. I have confirmed they are coming from the single reindex operation and not some other activity happening at the same time (or left over from earlier). Is this expected? What determines how many scroll contexts a reindex API call will use?
Specifically, I am calling the API like this:
curl -X POST <my_cluster>:9200/_reindex?slices=32 -H 'Content-Type: application/json' -d '
{
"dest": {
"index": <src_index>
},
"source": {
"index": <dest_index>,
"size": 5000,
"query": <my_query>
}
}
'