[Star Tree] [Search] Support for metric aggregations with/without term query

[sandeshkr419]

Disclaimer

These changes are built on top of unmerged/in-review indexing changes, Reviewers kindly ignore this commit while reviewing this change. When the depending changes are merged, will remove the Do not merge from title, and add chngelog to the PR - avoiding adding now to avoid unnecessary rebasing/conflicts.

Description

For an index supporting star-tree composite index, this changes tries to achieve resolving a metric aggregation with/without a numeric terms query with the help of star-tree.

In present state, the PR capture changes for sum, max, min, avg, value-count aggregation. Once the high level changes in sum aggregation are reviewed, will increment with other aggregations.

Approach

A new StarTreeQuery is introduced which helps resolve to star-tree documents. This star-tree query is formed (if it can be) at the shard level, this is not done at coordinator level to avoid node to node transportation. Also, all the information is present at shard level and OpenSearch does majority of query rewrite at shard level itself. This star tree query is encapsulated in an OriginalOrStarTreeQuery which helps preserve the original query alongwith the new star tree query. This encapsulation is done so as to preserve both the queries and decision whether to use which query can be taken at a segment level.

High Level Operations (to make code reviewing simple):

1. Parsing a Search Query, to decide whether or not to use star tree flow: SearchService.java, QueryShardContext.java
2. Parsed SearchQuery: StarTreeQuery.java, OriginalOrStarTreeQuery.java
3. Utilities for resolving star-tree given certain predicates (derived from term query of search request), and tree traversal - StarTreeFilter.java
4. Changes in 5 metric Aggregators - SumAggregator.java, MaxAggregator.java, MinAggregator.java, AvgAggregator.java, ValueCountAggregator.java
5. Supporting changes - changes in access level of aggregator factory classes so identify correct aggregations during request parsing.

TODO in this PR:

1. Test cases to be added.
2. Support for max, min, count, avg metric aggregations.

This PR depends on #14809, therefore the depending unmerged changes have been utilized for now in my private fork to discuss the changes in parallel.

Example query shape:

No query + Agg:

{
    "size": 0,
    "aggs": {
                        "sum_status": {
                            "sum": {
                                "field": "size"
                            }
                        }
                    }
}

Original Response:

{
    "took": 8378,
    "timed_out": false,
    "_shards": {
        "total": 1,
        "successful": 1,
        "skipped": 0,
        "failed": 0
    },
    "hits": {
        "total": {
            "value": 1009,
            "relation": "eq"
        },
        "max_score": null,
        "hits": []
    },
    "aggregations": {
        "sum_status": {
            "value": 615579.0
        }
    }
}

Star Tree Response:

{
    "took": 16394,
    "timed_out": false,
    "_shards": {
        "total": 1,
        "successful": 1,
        "skipped": 0,
        "failed": 0
    },
    "hits": {
        "total": {
            "value": 1,
            "relation": "eq"
        },
        "max_score": null,
        "hits": []
    },
    "aggregations": {
        "sum_status": {
            "value": 615579.0
        }
    }
}

Request:

{
    "query": {
        "term": {
            "status": 200
        }
    },
    "size": 0,
    "aggs": {
                        "sum_status": {
                            "sum": {
                                "field": "size"
                            }
                        }
                    }
}

Original Response:

{
    "took": 4038,
    "timed_out": false,
    "_shards": {
        "total": 1,
        "successful": 1,
        "skipped": 0,
        "failed": 0
    },
    "hits": {
        "total": {
            "value": 42,
            "relation": "eq"
        },
        "max_score": null,
        "hits": []
    },
    "aggregations": {
        "sum_status": {
            "value": 24745.0
        }
    }
}

Star Tree Flow Response:

{
    "took": 21120,
    "timed_out": false,
    "_shards": {
        "total": 1,
        "successful": 1,
        "skipped": 0,
        "failed": 0
    },
    "hits": {
        "total": {
            "value": 1,
            "relation": "eq"
        },
        "max_score": null,
        "hits": []
    },
    "aggregations": {
        "sum_status": {
            "value": 24745.0
        }
    }
}

Multi-aggs in a single request:

{ 
    "query": {
        "term": {
            "status": 201
        }
    },
    "size": 0,
    "aggs": {
                        "sum_status": {
                            "sum": {
                                "field": "size"
                            }
                        },
                        "max_status": {
                            "max": {
                                "field": "size"
                            }
                        },
                        "min_status": {
                            "min": {
                                "field": "size"
                            }
                        },
                        "avg_status": {
                            "avg": {
                                "field": "size"
                            }
                        },
                        "count_status": {
                            "value_count": {
                                "field": "size"
                            }
                        }

                    }
}

Star Tree Response:

{
    "took": 28,
    "timed_out": false,
    "_shards": {
        "total": 1,
        "successful": 1,
        "skipped": 0,
        "failed": 0
    },
    "hits": {
        "total": {
            "value": 3,
            "relation": "eq"
        },
        "max_score": null,
        "hits": []
    },
    "aggregations": {
        "max_status": {
            "value": 1000.0
        },
        "sum_status": {
            "value": 32154.0
        },
        "count_status": {
            "value": 53
        },
        "avg_status": {
            "value": 606.6792452830189
        },
        "min_status": {
            "value": 200.0
        }
    }
}

Unsupported Metric Operation via star-tree:

{ 
    "query": {
        "term": {
            "status": 201
        }
    },
    "size": 0,
    "aggs": {
                        "sum_status": {
                            "sum": {
                                "field": "size"
                            }
                        },
                        "max_status": {
                            "max": {
                                "field": "size"
                            }
                        },
                        "min_status": {
                            "min": {
                                "field": "size"
                            }
                        },
                        "avg_status": {
                            "avg": {
                                "field": "size"
                            }
                        },
                        "count_status": {
                            "stats": {
                                "field": "size"
                            }
                        }

                    }
}

Defaults to original code-flow (verified by hits reported):

{
    "took": 137433,
    "timed_out": false,
    "_shards": {
        "total": 1,
        "successful": 1,
        "skipped": 0,
        "failed": 0
    },
    "hits": {
        "total": {
            "value": 53,
            "relation": "eq"
        },
        "max_score": null,
        "hits": []
    },
    "aggregations": {
        "max_status": {
            "value": 1000.0
        },
        "sum_status": {
            "value": 32154.0
        },
        "count_status": {
            "count": 53,
            "min": 200.0,
            "max": 1000.0,
            "avg": 606.6792452830189,
            "sum": 32154.0
        },
        "avg_status": {
            "value": 606.6792452830189
        },
        "min_status": {
            "value": 200.0
        }
    }
}

Approach:

1. The query shape is identified at the shard level (SearchService.java) and the query/aggregation (if can be resolved by star-tree) is parsed to a star-tree query.
2. The star-tree query is wrapped around OriginalOrStarTreeQuery to preserve the original query - this is because the decision to decide which implementation (default/startree) to use can be taken for a segment level.
3. If star-tree can be utilized to answer the query, the star-tree document set is then collected by the relevant aggregator/collector. In this POC, I have made changes to SumAggregator to demonstrate the flow of changes.

Related Issues

#15257

Check List

• New functionality includes testing.
  • All tests pass
• New functionality has been documented.
  • New functionality has javadoc added
• Failing checks are inspected and point to the corresponding known issue(s) (See: Troubleshooting Failing Builds)
• Commits are signed per the DCO using --signoff
• Commit changes are listed out in CHANGELOG.md file (See: Changelog)
• 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.

[github-actions]

❌ Gradle check result for ad54ace: FAILURE

Please examine the workflow log, locate, and copy-paste the failure(s) below, then iterate to green. Is the failure a flaky test unrelated to your change?

[github-actions]

❌ Gradle check result for 6c6be02: FAILURE

Please examine the workflow log, locate, and copy-paste the failure(s) below, then iterate to green. Is the failure a flaky test unrelated to your change?

[github-actions]

❌ Gradle check result for 995db38: FAILURE

Please examine the workflow log, locate, and copy-paste the failure(s) below, then iterate to green. Is the failure a flaky test unrelated to your change?

[github-actions]

❌ Gradle check result for 885a383: FAILURE

Please examine the workflow log, locate, and copy-paste the failure(s) below, then iterate to green. Is the failure a flaky test unrelated to your change?

[github-actions]

❌ Gradle check result for 9491aae: FAILURE

Please examine the workflow log, locate, and copy-paste the failure(s) below, then iterate to green. Is the failure a flaky test unrelated to your change?

[github-actions]

❌ Gradle check result for a01c6e2: FAILURE

Please examine the workflow log, locate, and copy-paste the failure(s) below, then iterate to green. Is the failure a flaky test unrelated to your change?

[github-actions]

❌ Gradle check result for e4a270a: FAILURE

Please examine the workflow log, locate, and copy-paste the failure(s) below, then iterate to green. Is the failure a flaky test unrelated to your change?

[github-actions]

✅ Gradle check result for 89c845d: SUCCESS

[msfroh]

I guess you kind of have a check on the indexing side of things, since we currently only support dimensions defined over numeric fields, I think. So, the field wouldn't be present in supportedDimensions if it's not numeric. Eventually, we will want to support non-string values, though, right?

[msfroh]

Why do these field names start with underscore? That's kind of weird in Java.

[msfroh]

Can you please add a dedicated unit test for this class with some randomized testing?

It has a lot of uncovered lines, and this logic is key to making sure the whole thing works.

Also, it would be nice to make this class package-private, since it's only used to help StarTreeQueryHelper. Would it make sense to move StarTreeQueryHelper into this package? Or move this class into that package? It feels like they should be in the same package, since they're closely connected.

Also, I'm not sure I see any value in instantiating an object of this class. The caller has StarTreeValues and a set of predicateEvaluators. They want to pass both to this code to get back a FixedBitSet. Passing them in the constructor and immediately calling getStarTreeResult is allocating an object that will never be used for anything else.

In theory, this whole thing could be a couple of static methods in StarTreeQueryHelper. (That said, we should still have more unit test coverage, regardless of where it lands.)

[msfroh]

I would avoid the ConcurrentHashMap and either take the IndexReader as input and say starTreeValuesMap = new FixedBitSet[reader.leaves().size()] or you could take numLeaves as a parameter and say starTreeValuesMap = new FixedBitSet[numLeaves]. (I have no real preference either way.)

If cacheStarTreeValues is false, you just say starTreeValuesMap = null.

Then you could add a pair of methods:

FixedBitSet getStarTreeValues(LeafReaderContext ctx) {
  if (starTreeValuesMap != null) {
    return starTreeValuesMap[ctx.ord];
  }
  return null;
}

void setStarTreeValues(LeafReaderContext ctx, FixedBitSet values) {
  if (starTreeValuesMap != null) {
    starTreeValuesMap[ctx.ord] = values;
  }
}

Nothing needs to be volatile and no need for a ConcurrentHashMap.

[github-actions]

❌ Gradle check result for c7b70b0: FAILURE

Please examine the workflow log, locate, and copy-paste the failure(s) below, then iterate to green. Is the failure a flaky test unrelated to your change?