mirror of
https://github.com/hanzoai/datastore.git
synced 2026-09-26 13:55:07 +00:00
`prepare_hive_data.sh` ran its first statement under a 150s `timeout`; while the Hive metastore was still starting the `hive` CLI blocked on the connection and was killed (exit code 124), so only a handful of attempts fit into the integration-test budget before pytest's 900s timeout fired. This timed out `Integration tests (arm_binary, distributed plan, 3/4)` for both test_select_without_where and test_orc_split_skip_state_is_query_local, even though every other shard passed. Probe the metastore with a short-timeout trivial query and poll instead: a stuck probe is killed quickly and retried, while a successful one returns fast once the metastore is up, fitting many more connection attempts into the budget. The actual DDL / data load still runs with the generous timeout. If `timeout` is unavailable the probe is skipped, preserving the previous behaviour. https://github.com/ClickHouse/ClickHouse/pull/108094 Co-Authored-By: Claude Opus 4.8 (1M context) <[email protected]>