Fix rendezvous hashing in complex queries

[ianton-ru]

Changelog category (leave one):

• Bug Fix (user-visible misbehavior in an official stable release)

Changelog entry (a user-readable short description of the changes that goes to CHANGELOG.md):

Rendevous hashing was not work properly in complex queries

Documentation entry for user-facing changes

Rendevous hashing was not work properly in complex queries like

insert into test_insert select * from s3Cluster('cluster', 'http://minio1:9001/root/data/hivebig/key=**', 'minio', 'minio123', 'CSVWithNames') SETTINGS parallel_distributed_insert_select=1

[arthurpassos]

Can you please provide more context? Why it wasn't working, the root problem and how it was fixed?

[ianton-ru]

Can you please provide more context? Why it wasn't working, the root problem and how it was fixed?

ids_of_nodes does not fill in some cases, it was optional and nullopt was used. Than connection_to_files got size 1, and requests failed with exception No list of nodes inside Task Distributer.. Or worse, with number_of_current_replica >= 1 array connection_to_files used out of bound with possible segfault.

Now ids_of_nodes filled always. Initializing moved inside getTaskIteratorExtension.

[Enmk]

LGTM