Dremio 25.2: EVNT_DATE filter on UNION view not pushed down correctly to MySQL sources

1

Hi team,
I’m using Dremio version 25.2.0-202410241428100111-a963b970 and encountering an issue with filter pushdown behavior when querying a union of two MySQL-backed views.

I’ve created the following views:

  1. campaign_fct

SELECT EVNT_DATE, CAMPAIGN_NAME, DEL_CNT, SUB_CNT
FROM mysql7.Autopilot.CAMPAIGN_DATA_FACT
WHERE EVNT_DATE <= DATE_SUB(CURRENT_DATE, 2)

  1. campaign_trans

SELECT EVNT_DATE, CAMPAIGN_NAME, SUM(DEL_CNT), SUM(SUB_CNT)
FROM mysql7.Autopilot.CAMPAIGN_DATA_TRANS
WHERE EVNT_DATE > DATE_SUB(CURRENT_DATE, 2)
GROUP BY EVNT_DATE, CAMPAIGN_NAME

  1. campaign_common

SELECT * FROM “campaign_report_r&d”.“campaign_fct”
UNION ALL
SELECT * FROM “campaign_report_r&d”.“campaign_trans”

When I run the following query:

SELECT * FROM “campaign_report_r&d”.“campaign_common”
WHERE EVNT_DATE >= ‘2025-04-07’

22242fe4-11c7-4f59-8c39-068fc8a24d90-evnt-gtr-eql.zip (14.4 KB)

Only the filter is pushed to CAMPAIGN_DATA_TRANS, which is expected.

However, when I run this query:

SELECT * FROM “campaign_report_r&d”.“campaign_common”
WHERE EVNT_DATE = ‘2025-04-07’

e29f31ce-c3ec-4815-9954-bee02667a4e8-evnt-equals.zip (14.4 KB)

The EVNT_DATE filter gets pushed to both CAMPAIGN_DATA_FACT and CAMPAIGN_DATA_TRANS, which is not expected since CAMPAIGN_DATA_FACT is restricted to dates <= CURRENT_DATE - 2.

All tables and views are based on MySQL data sources.

Please find the attached query profiles for both queries for reference.

Question:
Why is the EVNT_DATE = ‘2025-04-07’ filter being pushed to both underlying sources in the union? Is this a known behavior or optimization limitation? How can we ensure filters are correctly pushed only to relevant branches of the union?

Thanks in advance!

2

Hi team,

Could you please provide me with an update on the issue mentioned above?

Thanks in Advance

3

Hello Team,

Kindly update.

Thanks,
Ajay Babu Maguluri.

4

Hi team,

kindly update on the issue mentioned above.

Thanks in Advance

5

@JoiceJacob @AjayBabuM At this point it looks like that is what the push down is doing. Do you have an instance where pushing down the filter to both sides of the union is causing a performance issue?