mirror of
https://github.com/hanzoai/datastore.git
synced 2026-09-26 13:55:07 +00:00
Alexey MilovidovandClaude Opus 4.7
ccbc12a160
Covers the scenario from the review on https://github.com/ClickHouse/ClickHouse/pull/100141: a two-shard cluster where the local shard has no underlying table and the remote shard has the table populated. With `skip_unavailable_shards = 1` and `prefer_localhost_replica = 1`, the local shard is skipped and the query reads from the remote shard. Without the setting, the local shard fails with `UNKNOWN_TABLE`. Co-Authored-By: Claude Opus 4.7 (1M context) <[email protected]>
56 lines
1.8 KiB
Python
56 lines
1.8 KiB
Python
import pytest
|
|
|
|
from helpers.client import QueryRuntimeException
|
|
from helpers.cluster import ClickHouseCluster
|
|
|
|
cluster = ClickHouseCluster(__file__)
|
|
|
|
node1 = cluster.add_instance("node1", main_configs=["configs/remote_servers.xml"])
|
|
node2 = cluster.add_instance("node2", main_configs=["configs/remote_servers.xml"])
|
|
|
|
|
|
@pytest.fixture(scope="module")
|
|
def start_cluster():
|
|
try:
|
|
cluster.start()
|
|
|
|
# Table exists only on the remote shard (node2). The local shard (node1)
|
|
# has no underlying table — querying it must throw UNKNOWN_TABLE unless
|
|
# skip_unavailable_shards lets us drop the local shard from the query.
|
|
node2.query(
|
|
"CREATE TABLE data (x UInt32) ENGINE = MergeTree() ORDER BY x"
|
|
)
|
|
node2.query("INSERT INTO data VALUES (1), (2), (3)")
|
|
|
|
node1.query(
|
|
"""
|
|
CREATE TABLE dist (x UInt32)
|
|
ENGINE = Distributed(local_and_remote, currentDatabase(), data)
|
|
"""
|
|
)
|
|
|
|
yield cluster
|
|
finally:
|
|
cluster.shutdown()
|
|
|
|
|
|
def test_skip_local_shard_when_table_missing(start_cluster):
|
|
# Without skip_unavailable_shards the local shard fails because `data` is
|
|
# absent on node1, and there are no other replicas to fall back to.
|
|
with pytest.raises(QueryRuntimeException, match="UNKNOWN_TABLE"):
|
|
node1.query(
|
|
"SELECT sum(x) FROM dist SETTINGS prefer_localhost_replica = 1"
|
|
)
|
|
|
|
# With skip_unavailable_shards the local shard is skipped and we read only
|
|
# from the remote shard (node2), which has the table populated.
|
|
assert (
|
|
node1.query(
|
|
"""
|
|
SELECT sum(x) FROM dist
|
|
SETTINGS skip_unavailable_shards = 1, prefer_localhost_replica = 1
|
|
"""
|
|
).strip()
|
|
== "6"
|
|
)
|