Prevent resetting latest flag of real-time when starting historical analysis

[kaituo]

Description

This PR addresses a bug where starting a historical analysis after a real-time analysis on the same detector caused the real-time task’s latest flag to be incorrectly reset to false by the historical run.

The fix ensures that only the latest flags of the same analysis type are reset:

• Real-time analysis will only reset the latest flag of previous real-time analyses.
• Historical analysis will only reset the latest flag of previous historical analyses.

Additional Changes:

• Set a minimum recencyEmphasis value of 2 to align with RCF requirements.
• Updated imputation logic to execute only when the entity maps to the local node ID in the hash ring, reducing errors when models of the same entity exist on multiple nodes.

Testing:

• Added an integration test to reproduce the bug and verified the fix.

Check List

• New functionality includes testing.
• New functionality has been documented.
• API changes companion pull request created.
• Commits are signed per the DCO using --signoff.
• Public documentation issue/PR created.

By submitting this pull request, I confirm that my contribution is made under the terms of the Apache 2.0 license.
For more information on following Developer Certificate of Origin and signing off your commits, please check here.

[codecov]

Codecov Report

Attention: Patch coverage is 53.33333% with 7 lines in your changes missing coverage. Please review.

Project coverage is 77.55%. Comparing base (3f0fc8c) to head (065a64b).
Report is 9 commits behind head on main.

Files with missing lines	Patch %	Lines
...search/ad/transport/ADHCImputeTransportAction.java	30.00%	0 Missing and 7 partials ⚠️

Additional details and impacted files

@@             Coverage Diff              @@
##               main    #1287      +/-   ##
============================================
+ Coverage     71.83%   77.55%   +5.72%     
- Complexity     4898     5432     +534     
============================================
  Files           518      532      +14     
  Lines         22879    23263     +384     
  Branches       2245     2305      +60     
============================================
+ Hits          16434    18041    +1607     
+ Misses         5410     4169    -1241     
- Partials       1035     1053      +18     

Flag	Coverage Δ
plugin	77.55% <53.33%> (+5.72%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

Files with missing lines	Coverage Δ
...ain/java/org/opensearch/ad/task/ADTaskManager.java	75.40% <100.00%> (-0.24%)	⬇️
.../opensearch/forecast/task/ForecastTaskManager.java	2.76% <ø> (ø)
...g/opensearch/timeseries/ml/RealTimeInferencer.java	90.14% <ø> (ø)
...n/java/org/opensearch/timeseries/model/Config.java	86.02% <100.00%> (-1.39%)	⬇️
...va/org/opensearch/timeseries/task/TaskManager.java	81.38% <100.00%> (+2.81%)	⬆️
...search/ad/transport/ADHCImputeTransportAction.java	69.76% <30.00%> (ø)

... and 105 files with indirect coverage changes

[amitgalitz]

before we could have occasions where two nodes have the same model for the same entity, so two nodes at the same time will find if there is an anomaly on the interval?

[kaituo]

yes.

[amitgalitz]

would this mean we would write result twice or we handled it later to only be written once?

[kaituo]

We would only write it once (not guaranteed) as here we only process imputation on a node the entity is mapped to.