[ 
https://issues.apache.org/jira/browse/SOLR-11676?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
 ]

Varun Thacker updated SOLR-11676:
---------------------------------
    Attachment: SOLR-11676.patch

Here's another approach to the fix which adds the validation in 
CollectionsHandler .

I'll add a test as well tomorrow

[~ab] I noticed that SimClusterStateProvider calls 
ClusterStateMutator#createCollection to mock collection creation. Today in 
CollectionsHandler.java we have some validation for the create command already, 
like verify rule syntax etc. Is it okay if we don't get the benefits or test 
coverage of the validation here?

> nrt replicas is always 1 when not specified
> -------------------------------------------
>
>                 Key: SOLR-11676
>                 URL: https://issues.apache.org/jira/browse/SOLR-11676
>             Project: Solr
>          Issue Type: Bug
>      Security Level: Public(Default Security Level. Issues are Public) 
>            Reporter: Varun Thacker
>         Attachments: SOLR-11676.patch, SOLR-11676.patch
>
>
> I created 1 2 shard X 2 replica collection . Here's the log entry for it
> {code}
> 2017-11-27 06:43:47.071 INFO  (qtp159259014-22) [   ] 
> o.a.s.h.a.CollectionsHandler Invoked Collection Action :create with params 
> replicationFactor=2&routerName=compositeId&collection.configName=_default&maxShardsPerNode=2&name=test_recovery&router.name=compositeId&action=CREATE&numShards=2&wt=json&_=1511764995711
>  and sendToOCPQueue=true
> {code}
> And then when I look at the state.json file I see nrtReplicas is set to 1. 
> Any combination of numShards and replicationFactor without explicitly 
> specifying the "nrtReplicas" param puts the "nrtReplicas" as 1 instead of 
> using the replicationFactor value
> {code}
> {"test_recovery":{
>     "pullReplicas":"0",
>     "replicationFactor":"2",
> ...
>     "nrtReplicas":"1",
>     "tlogReplicas":"0",
> ..
> {code}



--
This message was sent by Atlassian JIRA
(v6.4.14#64029)

---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to