[Bug]

Aberta
  • Algo errado?
#2,255 1 comentário 0 reações 0 responsáveis Ver no GitHub

Mantenedores costumam responder em até 2 dias

Ninguém assumiu esta issue ainda.

Avaliação

Dificuldade
4/5
Tempo estimado
3-5 dias
Facilidade para iniciantes
40/100
Tipo de issue
Bug
Clareza
Razoavelmente clara
Status de atividade
Ativa
Stack de tecnologia
c, postgresql
Domínio
backend, databases

Direção de pesquisa

Start with src/backend/gpopt/translate/CTranslatorQueryToDXL.cpp and the optimizer_enable_multiple_distinct_aggs definition in guc_gp.c; the issue identifies the guard that causes fallback. Check how the optimizer handles multiple distinct qualified aggregates and review existing tests for that behavior. Done means establishing whether enabling the option is supported and safe, with evidence about plan behavior and the reason it defaults off.

Escrita pelo modelo de indexação a partir do texto da issue.

Descrição

type: Bug
Apache Cloudberry version
Environment
  • Cloudberry version: <Ausgabe von SELECT version(); einfügen>
  • Segments: 4
  • optimizer = on
  • join_collapse_limit = 20, from_collapse_limit = 20 (defaults)
What happened
Environment
  • Cloudberry version: <Ausgabe von SELECT version(); einfügen>
  • Segments: 4
  • optimizer = on
  • join_collapse_limit = 20, from_collapse_limit = 20 (defaults)
Description

A BI tool (Yellowfin) generates queries with several COUNT(DISTINCT ...) on different columns.
GPORCA refuses these queries and falls back to the Postgres-based planner:

INFO:  GPORCA failed to produce a plan, falling back to planner
DETAIL:  Feature not supported: Multiple Distinct Qualified Aggregates are disabled in the optimizer

In our case the query joins ~20 tables (LEFT JOINs) with 7 COUNT(DISTINCT ...) and a wide GROUP BY.
After the fallback, the Postgres planner (with join_collapse_limit = 20) needs a very long time to plan:

Planning Time: 192018.556 ms
Optimizer: Postgres query optimizer
Execution Time: 50.880 ms

So the fallback itself turns a trivial query into a ~3 minute planning-only operation.

Questions
  1. Is this limitation still in place in the latest Cloudberry release, or is there a version where ORCA handles multiple DQAs by default?
  2. Is it safe/supported to set optimizer_enable_multiple_distinct_aggs = on in production (e.g. per role), given it is classified as a developer option? Are there known issues or plan-quality regressions (CXformGbAggWithMDQA2Join)?
  3. Is there a reason why the default is off (performance, correctness)?
  4. Are there recommendations to limit the planning time after an ORCA fallback for queries with many joins (e.g. lowering join_collapse_limit / from_collapse_limit for such roles)?

Thanks!

What you think should happen instead

No response

How to reproduce
Minimal reproduction
CREATE TABLE t_mdqa (a int, b int, c int) DISTRIBUTED BY (a);
SET optimizer = on;
SET optimizer_trace_fallback = on;
EXPLAIN SELECT count(DISTINCT a), count(DISTINCT b) FROM t_mdqa GROUP BY c;
-- -> falls back to Postgres-based planner

As far as I can see, the check is still present in current main
(src/backend/gpopt/translate/CTranslatorQueryToDXL.cpp:
if (1 < num_dqa && !optimizer_enable_multiple_distinct_aggs)),
and optimizer_enable_multiple_distinct_aggs is defined as a DEVELOPER_OPTIONS GUC in guc_gp.c.

Operating System

NAME="Red Hat Enterprise Linux" VERSION="8.10 (Ootpa)" ID="rhel" ID_LIKE="fedora" VERSION_ID="8.10" PLATFORM_ID="platform:el8" PRETTY_NAME="Red Hat Enterprise Linux 8.10 (Ootpa)" ANSI_COLOR="0;31" CPE_NAME="cpe:/o:redhat:enterprise_linux:8::baseos" HOME_URL="https://www.redhat.com/" DOCUMENTATION_URL="https://access.redhat.com/documentation/en-us/red_hat_enterprise_linux/8" BUG_REPORT_URL="https://issues.redhat.com/" REDHAT_BUGZILLA_PRODUCT="Red Hat Enterprise Linux 8" REDHAT_BUGZILLA_PRODUCT_VERSION=8.10 REDHAT_SUPPORT_PRODUCT="Red Hat Enterprise Linux" REDHAT_SUPPORT_PRODUCT_VERSION="8.10" Red Hat Enterprise Linux release 8.10 (Ootpa) Red Hat Enterprise Linux release 8.10 (Ootpa)

Anything else

No response

Are you willing to submit PR?
  • Yes, I am willing to submit a PR!
Code of Conduct
Linguagem predominante
C
Estrelas
1.4k
Forks
268
Merge médio
4d 3h
PRs com merge (30d)
78

Preparar o ambiente

Primeiros passos

  1. Leia a issue inteira e depois o guia de contribuição do projeto.
  2. Comente na issue dizendo que vai assumir — evita que duas pessoas façam o mesmo trabalho.
  3. Faça um fork do repositório e trabalhe em uma branch.
  4. Abra um pull request que referencie o número da issue.

Mais de apache/cloudberry

Todas as issues de apache/cloudberry

Issues semelhantes

Mais issues de C

Receba novas issues na sua caixa de entrada

Um resumo curto de issues do GitHub para quem está começando.