Geoff Genz
Geoff Genz
That sounds like an incompatibility in the elementary package. The data in a temporary (in memory) table on one node will not be available on other nodes in the cluster....
Fair point on the `temporary table` code. That long predates cluster functionality and is definitely a likely source of future bugs when dbt either runs in cluster mode or hits...
Thanks for the detailed report. How large is the dataset you're testing with? Is it possible to query the same data on an older ClickHouse version? That initial binary response...
Also do you get the same error when just calling `query` instead of `query_df`
Thanks for the additional information. I'm guessing this is some change in Native format related to the very large dataset and the later ClickHouse version, so I'll check with the...
Can you try one more thing? I want to ensure that the error is not compression related, so if you create the client with `compression: None`, does it still fail?
You can also try `compression: zstd` (lz4 is the default if nothing else is specified).
I have not been able to reproduce this locally despite very large datasets. There was a significant refactor of the ClickHouse HTTP server code a few months ago, so that...
Again thanks for the detailed investigation. If I understanding this right, you are getting compressed data even if there is not an `Accept-Encoding` header in the request header and no...
I'm still confused by one thing -- you said this worked if you used "no-compression" mode. What did you mean by that? If you set `compression=None` in the client that...