[Bug] Warm stage segments exceed configured segment_interval after lifecycle migration from hot

[mrproliu]

Search before asking

• I had searched in the issues and found no similar issues.

Apache SkyWalking Component

BanyanDB (apache/skywalking-banyandb)

What happened

After lifecycle migration from hot to warm stage, the actual segment time ranges on warm nodes exceed the configured segment_interval. For sw_metricsMinute:

Stage	Configured segment_interval
Hot	1 day
Warm	3 days

Data arrives day by day via lifecycle (no data gaps expected with 1-day hot interval). But the actual warm segments show:

Segment	Actual range	Duration	Expected
seg-20260325	03-25 ~ 03-27	2 days	3 days
seg-20260327	03-27 ~ 04-01	5 days	3 days
seg-20260401	04-01 ~ 04-04	3 days	3 days ✓

Both warm-0 and warm-1 show identical non-standard boundaries. Part metadata in seg-20260327 confirms it contains 5 full days of data:

Date	Rows	Notes
03-27	~6,230,000	Normal
03-28	~6,213,000	Normal
03-29	~6,204,000	Normal
03-30	~6,191,000	In compacted part spanning 03-29 ~ 03-30
03-31	~5,865,000	Normal
Total	30,703,391

Since the hot stage uses a 1-day segment_interval, the lifecycle migrates data on a day-by-day basis. Every day has continuous data (~6M rows), so there should be no data gaps that would justify a segment exceeding the
configured 3-day interval on warm.

What you expected to happen

Warm stage segments should respect the configured segment_interval of 3 days:

Segment	Expected range	Duration
seg-20260326	03-26 ~ 03-29	3 days
seg-20260329	03-29 ~ 04-01	3 days
seg-20260401	04-01 ~ 04-04	3 days

How to reproduce

Hot and warm stages have different segment_interval (e.g., hot = 1 day, warm = 3 days).

Anything else

No response

Are you willing to submit a pull request to fix on your own?

• Yes I am willing to submit a pull request on my own!

Code of Conduct

• I agree to follow this project's Code of Conduct