Hi all,
we do have some trouble with inconsistent search results from collections that
have a single shard and 2 replicas - Solr 8.11.4
The collections use a field with following field type:
<fieldType name="text_special_engrm_sku" class="solr.TextField"
positionIncrementGap="100">
<analyzer type="index">
<tokenizer class="solr.KeywordTokenizerFactory"/>
<filter class="solr.WordDelimiterGraphFilterFactory"
generateWordParts="0" generateNumberParts="0" splitOnCaseChange="0"
splitOnNumerics="0" catenateWords="1" catenateNumbers="1" catenateAll="1"/>
<filter class="solr.EdgeNGramFilterFactory" minGramSize="2"
maxGramSize="18"/>
<filter class="solr.LowerCaseFilterFactory"/>
<filter class="solr.FlattenGraphFilterFactory"/> <!-- required on index
analyzers after graph filters -->
</analyzer>
<analyzer type="query">
<tokenizer class="solr.KeywordTokenizerFactory"/>
<filter class="solr.WordDelimiterGraphFilterFactory"
generateWordParts="0" generateNumberParts="0" splitOnCaseChange="0"
splitOnNumerics="0" catenateWords="1" catenateNumbers="1" catenateAll="1"/>
<filter class="solr.LowerCaseFilterFactory"/>
</analyzer>
</fieldType>
and is used to index&search kinds of combined letter/numbers SKUs e.g.
"XY543-321" and to be able to search for e.g. "XY"
On occasion the search stops finding documents when searching for SKUs starting
with letters, but only when the search is executed at one (the affected)
replica. The other replica is returning results as expected.
We inspected the terms of the field with Solr Admin UI or the luke request
handler /solr/affected_replica_core/admin/luke?&wt=json&fl=sku and found a
difference in the distinct term count and upper case terms like "XY" in the
affected replica.
So sometimes lower casing seems to be not applied in one of the replicas.
Currently we do not know why & with what circumstances this happens. Anyone
experienced similar issues? Ideas are welcome.
Kind regards
Steffen
