Fix OpenSearch bucket aggregations (#272)

* Fix OpenSearch bucket aggregations

In order to build filter lists, we collect the values assigned to each key
using an OpenSearch bucket aggregation. Unfortunately, by default, a bucket
aggregation returns only the first 10 buckets; in at least one specific case
where someone noticed (the releaseStream keyword) we've exceeded 10 values and
expected terms aren't being reported.

This uses the `field`'s `size` keyword to return up to 1000 buckets instead.
This is likely overkill for many cases, but should be "safe" for most cases.

I considered adding code into `search.py` to analyzed returned bucket
aggregations for a non-zero `sum_other_doc_count` (which means we could have
returned additional terms), and report any case where this happens for log
stream analysis. This doesn't solve any additional problems, however, adds
more code, and is a little tricky given the logic to merge aggregations
against two OpenSearch instances. If some reviewer wants to insist, or if I
start feeling too guilty before I get an approval, I'll go ahead and do it.

Closes #271

Author: dbutenhof

backend/app/api/v1/commons/constants.py

@@ -29,6 +29,7 @@
     "formal",
     "ciSystem",
 )
+AGG_BUCKET_SIZE = 1000
 MAX_PAGE = 10000
 OCP_SHORT_VER_LEN = 6
 

backend/app/api/v1/commons/utils.py

@@ -124,7 +124,7 @@ def normalize_pagination(offset: Optional[int], size: Optional[int]) -> tuple[in
 def buildAggregateQuery(constant_dict):
     aggregate = {}
     for x, y in constant_dict.items():
-        obj = {x: {"terms": {"field": y}}}
+        obj = {x: {"terms": {"field": y, "size": constants.AGG_BUCKET_SIZE}}}
         aggregate.update(obj)
     return aggregate
 

backend/app/api/v1/endpoints/ocp/graph.py

@@ -3,6 +3,7 @@
 from fastapi import APIRouter
 import pandas as pd
 
+from app.api.v1.commons.constants import AGG_BUCKET_SIZE
 from app.api.v1.commons.utils import getMetadata
 from app.services.search import ElasticService
 
@@ -381,11 +382,11 @@ async def getBurnerCPUResults(uuids: list, namespace: str, index: str):
         "size": 0,
         "aggs": {
             "time": {
-                "terms": {"field": "uuid.keyword"},
+                "terms": {"field": "uuid.keyword", "size": AGG_BUCKET_SIZE},
                 "aggs": {"time": {"avg": {"field": "timestamp"}}},
             },
             "uuid": {
-                "terms": {"field": "uuid.keyword"},
+                "terms": {"field": "uuid.keyword", "size": AGG_BUCKET_SIZE},
                 "aggs": {"cpu": {"avg": {"field": "value"}}},
             },
         },

backend/app/api/v1/endpoints/quay/quayGraphs.py

@@ -1,5 +1,6 @@
 from fastapi import APIRouter
 
+from app.api.v1.commons.constants import AGG_BUCKET_SIZE
 from app.api.v1.commons.utils import getMetadata
 from app.services.search import ElasticService
 
@@ -157,7 +158,7 @@ async def getImageMetrics(uuids: list, index: str):
         "size": 0,
         "aggs": {
             "uuid": {
-                "terms": {"field": "uuid.keyword"},
+                "terms": {"field": "uuid.keyword", "size": AGG_BUCKET_SIZE},
                 "aggs": {
                     "latency": {"avg": {"field": "elapsed_time"}},
                     "success_count": {"sum": {"field": "success_count"}},

backend/tests/unit/test_utils.py

@@ -423,9 +423,9 @@ def test_build_aggregate_query_structure(self):
 
         expected_result = {
             "aggregate_structure": {
-                "platform": {"terms": {"field": "platform.keyword"}},
-                "benchmark": {"terms": {"field": "benchmark.keyword"}},
-                "jobStatus": {"terms": {"field": "jobStatus.keyword"}},
+                "platform": {"terms": {"field": "platform.keyword", "size": 1000}},
+                "benchmark": {"terms": {"field": "benchmark.keyword", "size": 1000}},
+                "jobStatus": {"terms": {"field": "jobStatus.keyword", "size": 1000}},
             },
             "correct_structure": True,
         }