[
https://issues.apache.org/jira/browse/SOLR-9069?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15271818#comment-15271818
]
David Smiley commented on SOLR-9069:
------------------------------------
BaseDistributedSearchTestCase actually doesn't bother me, it's the other
abstract classes below it for ZK -- AbstractDistribZkTestBase and
AbstractFullDistribZkTestBase. BaseDistributedSearchTestCase is very useful
for testing parity between sharded and non-sharded. Do you have plans to
introduce another mechanism to replace this key nature of
BaseDistributedSearchTestCase?
bq. ...migrate everything to SolrCloudTestCase. I'm going to be working on that
this week...
That sounds like a lot of work! You might first introduce a test shim,
deprecated from the outset, that exists to house many of the utility methods
involved. And then tests below AbstractDistribZkTestBase can subclass _this_
one. This will make the change less disruptive and more manageable.
> Cleanup/rename confusion of shardCount in tests actually being serverCount
> --------------------------------------------------------------------------
>
> Key: SOLR-9069
> URL: https://issues.apache.org/jira/browse/SOLR-9069
> Project: Solr
> Issue Type: Test
> Components: Tests
> Reporter: David Smiley
>
> BaseDistributedSearchTestCase has a shardCount field, which can be set
> directly or via the {{\@ShardsFixed}} annotation. For some (esp. older)
> tests, this is in fact the number of "shards" (disjoint slices of your
> overall data), but it's also the number of Solr/Jetty servers/nodes.
> Subclasses like AbstractFullDistribZkTestBase define sliceCount, adding to
> the confusion, and define however many shards (slices) they want --
> separately from the number of servers (which is configured confusingly, as
> stated, via ShardsFixed). This is confusing! I'm not 100% sure what the
> solution is, I'll suggest some ideas, but we should discuss.
> Of course we got to this situation historically before SolrCloud existed; no
> excuses needed. Now we should fix it.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
