Using the terms component in Solr Cloud gives random result

[Markus Kalkbrenner <ma...@biologis.com>]

Hi,

I tried to use the terms component with the techproducts example in cloud mode and was surprised, that the results toggle for each request.

The response to
http://localhost:8983/solr/techproducts/terms?terms=true&terms.fl=name
permanently toggles between these two results:

{
  "responseHeader":{
    "zkConnected":true,
    "status":0,
    "QTime":0},
  "terms":{
    "name":[
      "184",3,
      "1gb",3,
      "3200",3,
      "400",3,
      "ddr",3,
      "memory",3,
      "pc",3,
      "pin",3,
      "sdram",3,
      "system",3]}}

{
  "responseHeader":{
    "zkConnected":true,
    "status":0,
    "QTime":0},
  "terms":{
    "name":[
      "one",3,
      "canon",2,
      "drive",2,
      "gb",2,
      "hard",2,
      "11",1,
      "133",1,
      "19",1,
      "250",1,
      "300",1]}}


To reproduce the issue, here’s how I setup the test environment:

bin/solr start -e cloud -noprompt

bin/solr create -c techproducts -s 2 -rf 2 -d server/solr/configsets/sample_techproducts_configs/conf -n sample_techproducts_configs

bin/post -c techproducts example/exampledocs/*.xml


In the admin UI I can see that 18 documents are indexed in one core and 14 in the other. (The techproducts example contains 32 docs in total.)

Obviously I expected to get the terms for all documents stored in the collection and not randomly from one core or the other.

Am I doing something wrong?

Regards,
Markus

Re: Using the terms component in Solr Cloud gives random result

[Markus Kalkbrenner <ma...@biologis.com>]

I’ll answer my own question:

setting distrib=true solved the issue … mostly

Our client requests wt=json&json.nl <http://json.nl/>=flat

But the result isn’t flat!

As soon I set distrib=true the json response is formatted as a map instead go being flat.

Did I find a bug or this a known limitation?

Markus

> Am 31.01.2019 um 19:43 schrieb Markus Kalkbrenner <ma...@biologis.com>:
> 
> Hi,
> 
> I tried to use the terms component with the techproducts example in cloud mode and was surprised, that the results toggle for each request.
> 
> The response to
> http://localhost:8983/solr/techproducts/terms?terms=true&terms.fl=name
> permanently toggles between these two results:
> 
> {
>  "responseHeader":{
>    "zkConnected":true,
>    "status":0,
>    "QTime":0},
>  "terms":{
>    "name":[
>      "184",3,
>      "1gb",3,
>      "3200",3,
>      "400",3,
>      "ddr",3,
>      "memory",3,
>      "pc",3,
>      "pin",3,
>      "sdram",3,
>      "system",3]}}
> 
> {
>  "responseHeader":{
>    "zkConnected":true,
>    "status":0,
>    "QTime":0},
>  "terms":{
>    "name":[
>      "one",3,
>      "canon",2,
>      "drive",2,
>      "gb",2,
>      "hard",2,
>      "11",1,
>      "133",1,
>      "19",1,
>      "250",1,
>      "300",1]}}
> 
> 
> To reproduce the issue, here’s how I setup the test environment:
> 
> bin/solr start -e cloud -noprompt
> 
> bin/solr create -c techproducts -s 2 -rf 2 -d server/solr/configsets/sample_techproducts_configs/conf -n sample_techproducts_configs
> 
> bin/post -c techproducts example/exampledocs/*.xml
> 
> 
> In the admin UI I can see that 18 documents are indexed in one core and 14 in the other. (The techproducts example contains 32 docs in total.)
> 
> Obviously I expected to get the terms for all documents stored in the collection and not randomly from one core or the other.
> 
> Am I doing something wrong?
> 
> Regards,
> Markus
> 
> 
> 

Markus Kalkbrenner
Dipl.-Ing. (FH) techn. Informatik
CTO

T +49 69 / 348 739 110

markus.kalkbrenner@biologis.com


bio.logis Genetic Information Management GmbH
Zentrale:
Olof-Palme-Str. 15
D-60439 Frankfurt am Main

Geschäftsführung: Prof. Dr. med. Daniela Steinberger, Dipl. Betriebswirt Enrico Just
Firmensitz: Altenhöferallee 3, 60438 Frankfurt am Main
Registergericht Frankfurt am Main, HRB 97945