aws / aws/amazon-redshift-python-driver
[Bug] BrokenPipe: server socket closed error when we execute any dbt model in dbt core with redshift adapter
- 主要语言
- Python
- 星标
- 220
- 派生
- 86
- PR 合并指标
- 30 天内没有已合并 PR
描述
We are facing an issue if we execute a dbt model in dbt core with redshift adapter in multi nodes or multi threads. We did some analysis to find the reason for a failure by checking the dbt logs and we found that at some random point one of the threads was going to idle state for a certain period of time and it was not executing any tasks sent to it and eventually it got failed with the below statement.
BrokenPipe: server socket closed. Please check that client side networking configurations such as Proxies, firewalls, VPN, etc. are not affecting your network connection.
We suspect this could be because of introduction of redshift_connector in dbt-redshift adapter by replacing psycopg2. We weren't facing any such issues in dbt older versions which used psycopg2.
Attaching the context from dbt log file
```
13:28:59.394786 [debug] [Thread-1 (]: Elementary: [dbt_tests] Artifacts changed.
.
.
.
13:53:01.835859 [debug] [Thread-1 (]: Elementary: Deleting from and inserting to: "elementary"."dbt_tests"
13:53:01.859812 [debug] [Thread-1 (]: Using redshift connection "model.elementary.dbt_tests"
13:53:01.860853 [debug] [Thread-1 (]: On model.elementary.dbt_tests: /* {"app": "dbt", "dbt_version": "1.8.9", "profile_name": "", "target_name": "", "node_id": "model.elementary.dbt_tests"} */
13:53:01.863164 [debug] [Thread-1 (]: Redshift adapter: Rolling back transaction.
13:53:01.864028 [debug] [Thread-1 (]: On model.elementary.dbt_tests: ROLLBACK
13:53:01.866800 [debug] [Thread-1 (]: Failed to rollback 'model.elementary.dbt_tests'
13:53:01.873315 [debug] [Thread-1 (]: On model.elementary.dbt_tests: Close
13:53:01.879586 [debug] [Thread-1 (]: Runtime Error in model dbt_tests (models/edr/dbt_artifacts/dbt_tests.sql)
BrokenPipe: server socket closed. Please check that client side networking configurations such as Proxies, firewalls, VPN, etc. are not affecting your network connection.
13:53:01.881652 [debug] [Thread-1 (]: Sending event: {'category': 'dbt', 'action': 'run_model', 'label': '382ae431-f242-4924-a625-8b4dd918ff58', 'context': []}
13:53:01.882856 [error] [Thread-1 (]: 1 of 1 ERROR creating sql incremental model elementary.dbt_tests .......... [ERROR in 1646.81s]
```
As you can see last time when connection was used in Thread-1 was at 13:28:59.394786 (first line)
After almost 25mins dbt got back to this connection to sent the next statement but the connection was no longer open
There is no keep_alive option in redshift_connector - this is something we had specified in the old redshift adapter.
贡献指南
调研方向
No repository files, tests, or entry points are named. Start by reproducing a multi-threaded dbt model run with the Redshift adapter and compare the connection lifecycle around the idle period shown in the logs; done means identifying and verifying the cause of the BrokenPipe failure.
由索引模型根据 Issue 内容生成。
评估
- 技术栈
- python
- 领域
- databases
- Issue 类型
- 缺陷
- 难度
- 4/5
- 预计耗时
- 3-5 天
- 活跃度
- 停滞
- 描述清晰度
- 需要澄清
- 新手友好度
- 25/100