[SOLR-6700] ChildDocTransformer doesn't return correct children after updating and optimising solr index

• Type: Bug
• Status: Closed
• Priority: Critical
• Resolution: Not A Bug
• Fix Version/s: 4.10.5
• Component/s: update
• Assignee: Unassigned
• Reporter: Bogdan Marinescu
• Created: 04/Nov/14 09:37
• Updated: 21/Aug/18 15:51
• Resolved: 21/Aug/18 15:51

Description

I have an index with nested documents.

schema.xml snippet

 <field name="id" type="string" indexed="true" stored="true" required="true" multiValued="false" />
<field name="entityType" type="int" indexed="true" stored="true" required="true"/>
<field name="pName" type="string" indexed="true" stored="true"/>
<field name="cAlbum" type="string" indexed="true" stored="true"/>
<field name="cSong" type="string" indexed="true" stored="true"/>
<field name="_root_" type="string" indexed="true" stored="true"/>
<field name="_version_" type="long" indexed="true" stored="true"/>

Afterwards I add the following documents:

<add>
  <doc>
    <field name="id">1</field>
    <field name="pName">Test Artist 1</field>
    <field name="entityType">1</field>
    <doc>
        <field name="id">11</field>
        <field name="cAlbum">Test Album 1</field>
	    <field name="cSong">Test Song 1</field>
        <field name="entityType">2</field>
    </doc>
  </doc>
  <doc>
    <field name="id">2</field>
    <field name="pName">Test Artist 2</field>
    <field name="entityType">1</field>
    <doc>
        <field name="id">22</field>
        <field name="cAlbum">Test Album 2</field>
	    <field name="cSong">Test Song 2</field>
        <field name="entityType">2</field>
    </doc>
  </doc>
</add>

After performing the following query

http://localhost:8983/solr/collection1/select?q=%7B!parent+which%3DentityType%3A1%7D&fl=*%2Cscore%2C%5Bchild+parentFilter%3DentityType%3A1%5D&wt=json&indent=true

I get a correct answer (child matches parent, check root field)

add docs

{
  "responseHeader":{
    "status":0,
    "QTime":1,
    "params":{
      "fl":"*,score,[child parentFilter=entityType:1]",
      "indent":"true",
      "q":"{!parent which=entityType:1}",
      "wt":"json"}},
  "response":{"numFound":2,"start":0,"maxScore":1.0,"docs":[
      {
        "id":"1",
        "pName":"Test Artist 1",
        "entityType":1,
        "_version_":1483832661048819712,
        "_root_":"1",
        "score":1.0,
        "_childDocuments_":[
        {
          "id":"11",
          "cAlbum":"Test Album 1",
          "cSong":"Test Song 1",
          "entityType":2,
          "_root_":"1"}]},
      {
        "id":"2",
        "pName":"Test Artist 2",
        "entityType":1,
        "_version_":1483832661050916864,
        "_root_":"2",
        "score":1.0,
        "_childDocuments_":[
        {
          "id":"22",
          "cAlbum":"Test Album 2",
          "cSong":"Test Song 2",
          "entityType":2,
          "_root_":"2"}]}]
  }}

Afterwards I try to update one document:

update doc

<add>
<doc>
<field name="id">1</field>
<field name="pName" update="set">INIT</field>
</doc>
</add>

After performing the previous query I get the right result (like the previous one but with the pName field updated).

The problem only comes after performing an optimize.
Now, the same query yields the following result:

{
  "responseHeader":{
    "status":0,
    "QTime":1,
    "params":{
      "fl":"*,score,[child parentFilter=entityType:1]",
      "indent":"true",
      "q":"{!parent which=entityType:1}",
      "wt":"json"}},
  "response":{"numFound":2,"start":0,"maxScore":1.0,"docs":[
      {
        "id":"2",
        "pName":"Test Artist 2",
        "entityType":1,
        "_version_":1483832661050916864,
        "_root_":"2",
        "score":1.0,
        "_childDocuments_":[
        {
          "id":"11",
          "cAlbum":"Test Album 1",
          "cSong":"Test Song 1",
          "entityType":2,
          "_root_":"1"},
        {
          "id":"22",
          "cAlbum":"Test Album 2",
          "cSong":"Test Song 2",
          "entityType":2,
          "_root_":"2"}]},
      {
        "id":"1",
        "pName":"INIT",
        "entityType":1,
        "_root_":"1",
        "_version_":1483832916867809280,
        "score":1.0}]
  }}

As can be seen, the document with id:2 now contains the child with id:11 that belongs to the document with id:1.

I haven't found any references on the web about this except http://blog.griddynamics.com/2013/09/solr-block-join-support.html
Similar issue: SOLR-6096

Is this problem known? Is there a workaround for this?

Issue Links

• duplicates: SOLR-5211 updating parent as childless makes old children orphans
• is related to: SOLR-12638 Support atomic updates of nested/child documents for nested-enabled schema
• is related to: SOLR-5142 Block Indexing / Join Improvements
• is related to: SOLR-10144 redesign block-join support
• relates to: SOLR-6596 Atomic update and adding child doc not working together