[jira] [Commented] (SOLR-7734) MapReduce Indexer can error when using collection

["Gregory Chanan (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/SOLR-7734?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14700579#comment-14700579 ] 

Gregory Chanan commented on SOLR-7734:
--------------------------------------

Looks good, I really like the new test.  All of my previous comments seem to be addressed.  Just a few minor issues/comments below:

Issue #1:
When I try "ant test" I get:
{code}
   [junit4]    > Throwable #1: java.lang.RuntimeException: Suite class org.apache.solr.hadoop.MiniMRTest should be a concrete class (not abstract).
{code}

Issue #2:
{code}
public void useSolrHomeDir() throws Exception {
    String[] prepend = {"--solr-home-dir=" + DROPALL_CONF_DIR.getAbsolutePath()};
{code}
you can't actually tell if this is going to zk or not.  Maybe overwrite zk with the "good" version or something beforehand?

Issue #3:
{code}
solr/contrib/morphlines-core/src/test-files/solr/dropall/conf/stopwords.txt 	
{code}
Do we need this file?  I don't see it referenced.

Issue #4:
{code}
    try (InputStream source = MapReduceIndexerTool.class.getResourceAsStream("/solrconfig.indexer.xml");
              FileOutputStream destination = new FileOutputStream(getSolrConfig(tmpSolrHomeDir))) {
            ByteStreams.copy(source, destination);
          }
          LOG.debug("Replaced zookeeper's solrconfig.xml with embedded version.");
{code}
This spacing looks funky here.

Issue #5:
{code}
SolrConfigMRTest
{code}
can you put the license first?  all the other test have the license first (or after the package).  I don't know if this fails the rat check or not, but seems good to be consistent.



> MapReduce Indexer can error when using collection
> -------------------------------------------------
>
>                 Key: SOLR-7734
>                 URL: https://issues.apache.org/jira/browse/SOLR-7734
>             Project: Solr
>          Issue Type: Bug
>          Components: contrib - MapReduce
>    Affects Versions: 5.2.1
>            Reporter: Mike Drob
>            Assignee: Gregory Chanan
>             Fix For: Trunk, 5.4
>
>         Attachments: SOLR-7734.patch, SOLR-7734.patch, SOLR-7734.patch, SOLR-7734.patch, SOLR-7734.patch
>
>
> When running the MapReduceIndexerTool, it will usually pull a {{solrconfig.xml}} from ZK for the collection that it is running against. This can be problematic for several reasons:
> * Performance: The configuration in ZK will likely have several query handlers, and lots of other components that don't make sense in an indexing-only use of EmbeddedSolrServer (ESS).
> * Classpath Resources: If the Solr services are using some kind of additional service (such as Sentry for auth) then the indexer will not have access to the necessary configurations without the user jumping through several hoops.
> * Distinct Configuration Needs: Enabling Soft Commits on the ESS doesn't make sense. There's other configurations that 
> * Update Chain Behaviours: I'm under the impression that UpdateChains may behave differently in ESS than a SolrCloud cluster. Is it safe to depend on consistent behaviour here?



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@lucene.apache.org
For additional commands, e-mail: dev-help@lucene.apache.org