[
https://issues.apache.org/jira/browse/SOLR-5865?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13936239#comment-13936239
]
Mark Miller commented on SOLR-5865:
-----------------------------------
Looks great!
+ // We could upload the minimum set of files rather than the directory,
but that requires keeping the list up to date
+ ZkController.uploadToZK(zkClient, new File(configDir),
ZkController.CONFIGS_ZKNODE + "/" + configName);
The main reason most of the cloud tests have gone with specifying which config
files to put in zk was that uploading the entire directory of test configs was
damn slow and then repeated for all cloud tests.
A better solution at some point would be a new test config folder just for
solrcloud. We already have a lot of configs, but we could probably merge some
things into this - like the common solrconfig and schema that almost all cloud
tests use anyway. If we kept it to one set, I think it would be an improvement
for cloud tests.
> Provide a MiniSolrCloudCluster to enable easier testing
> -------------------------------------------------------
>
> Key: SOLR-5865
> URL: https://issues.apache.org/jira/browse/SOLR-5865
> Project: Solr
> Issue Type: Improvement
> Components: SolrCloud
> Affects Versions: 4.7, 5.0
> Reporter: Gregory Chanan
> Attachments: SOLR-5865.patch
>
>
> Today, the SolrCloud tests are based on the LuceneTestCase class hierarchy,
> which has a couple of issues around support for downstream projects:
> - It's difficult to test SolrCloud support in a downstream project that may
> have its own test framework. For example, some projects have support for
> different storage backends (e.g. Solr/ElasticSearch/HBase) and want tests
> against each of the different backends. This is difficult to do cleanly,
> because the Solr tests require derivation from LuceneTestCase, while the
> other don't
> - The LuceneTestCase class hierarchy is really designed for internal solr
> tests (e.g. it randomizes a lot of parameters to get test coverage, but a
> downstream project probably doesn't care about that). It's also quite
> complicated and dense, much more so than a downstream project would want.
> Given these reasons, it would be nice to provide a simple
> "MiniSolrCloudCluster", similar to how HDFS provides a MiniHdfsCluster or
> HBase provides a MiniHBaseCluster.
--
This message was sent by Atlassian JIRA
(v6.2#6252)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
