ClickHouse build check — 20/21 builds are OK

An error in infamous "Sandbox" infrastructure.

How does this affect zookeeper usage and performance? Currently we have ClickHouse in multiple datacenters, but we use the replicated_can_become_leader to ensure all leaders are in the same datacenter as the "voting" zookeeper cluster. Before we did that we saw issues with too many active merges. Is the replicated_can_become_leader setting deprecated? Will this affect merge operations positively or negatively in our configuration?

Also, do we have to upgrade all replicas to 20.5 at the same time?

Is the replicated_can_become_leader setting deprecated?

The setting remains and has the same effect.
We allow multiple leaders but you can disable leadership for specific replicas with replicated_can_become_leader.

Also, do we have to upgrade all replicas to 20.5 at the same time?

No, it's 100% compatible. New versions can become leader if there is no old leader.

How does this affect zookeeper usage and performance?

The number of request to ZooKeeper may increase but insignificantly in most configurations.
Significant increase of ZooKeeper load may occur if you have very high number of replicas in a single shard.

Will this affect merge operations positively or negatively in our configuration?

Merges should work in the same way.
Also it can eliminate the need of replicated_can_become_leader tuning, but it's Ok to keep it as it was.

Thanks for the details!