risingwave_stream::executor::aggregation::agg_group: bad row count group=OwnedRow(...)

[kwannoel]

Some users recently encounter this, after cluster recovery.

thread 'risingwave-streaming-actor' panicked at /risingwave/src/stream/src/executor/aggregation/agg_group.rs:320:13: row count should be non-negative

[kwannoel]

Related #14031

[kwannoel]

CR: @hzxa21 who initially found that it could potentially lead to backwards compat issues.

I can confirm this should probably be due to case backwards compatibility.

On v1.4-rc:

drop table t cascade;
create table t(x boolean, a int, y boolean, b int);
insert into t values (true, 1, true, 2);
create materialized view m1 as select 'abc' as v1, case when x then a when y then b end as case_col from t group by case_col;
select * from m1;

On v1.5.0-rc:

delete from t;
flush;

Agg panics with:

thread 'risingwave-streaming-actor' panicked at /Users/noelkwan/projects/risingwave/src/stream/src/executor/aggregation/agg_group.rs:320:13:
row count should be non-negative
stack backtrace:
   0: rust_begin_unwind
             at /rustc/249624b5043013d18c00f0401ca431c1a6baa8cd/library/std/src/panicking.rs:597:5
   1: core::panicking::panic_fmt
             at /rustc/249624b5043013d18c00f0401ca431c1a6baa8cd/library/core/src/panicking.rs:72:14
   2: risingwave_stream::executor::aggregation::agg_group::AggGroup<S,Strtg>::curr_row_count
             at ./src/stream/src/executor/aggregation/agg_group.rs:320:13
   3: risingwave_stream::executor::aggregation::agg_group::AggGroup<S,Strtg>::apply_chunk::{{closure}}
             at ./src/stream/src/executor/aggregation/agg_group.rs:351:12
   4: risingwave_stream::executor::hash_agg::HashAggExecutor<K,S>::apply_chunk::{{closure}}
             at ./src/stream/src/executor/hash_agg.rs:419:18
   5: risingwave_stream::executor::hash_agg::HashAggExecutor<K,S>::execute_inner::{{closure}}
             at ./src/stream/src/executor/hash_agg.rs:650:68
   6: <futures_async_stream::try_stream::GenTryStream<G> as futures_core::stream::Stream>::poll_next
             at /Users/noelkwan/.cargo/registry/src/index.crates.io-6f17d22bba15001f/futures-async-stream-0.2.9/src/lib.rs:506:33
   7: <core::pin::Pin<P> as futures_core::stream::Stream>::poll_next
             at /Users/noelkwan/.cargo/registry/src/index.crates.io-6f17d22bba15001f/futures-core-0.3.28/src/stream.rs:120:9
   8: <core::pin::Pin<P> as futures_core::stream::Stream>::poll_next
             at /Users/noelkwan/.cargo/registry/src/index.crates.io-6f17d22bba15001f/futures-core-0.3.28/src/stream.rs:120:9
   9: futures_util::stream::stream::StreamExt::poll_next_unpin
             at /Users/noelkwan/.cargo/registry/src/index.crates.io-6f17d22bba15001f/futures-util-0.3.28/src/stream/stream/mod.rs:1632:9
Sat Dec 23 03:00:56 UTC 2023 [risedev]: Program exited with 139

[stdrc]

Is this possibly caused by #13890? Cc @wangrunji0408

[kwannoel]

Is this possibly caused by #13890? Cc @wangrunji0408

Yeah, may look at bugens comment in that PR

[MrCroxx]

Is this possibly caused by #13890? Cc @wangrunji0408

Yeah, may look at bugens comment in that PR

FYI: #13890 (comment)

[kwannoel]

One of the root cause is case expression. For the other root cause it is unclear.

For case expression, we can't really avoid the breaking change in expr framework.

We have to inform users of the breaking change via the release notes, and get them to recreate any MVs with case expressions. The v1.5 release notes have been updated for CASE expression.

https://docs.risingwave.com/release-notes/