[
https://issues.apache.org/jira/browse/SOLR-8127?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=18082691#comment-18082691
]
Lucas Kot-Zaniewski commented on SOLR-8127:
-------------------------------------------
[~dsmiley] From my profiling this seems to be caused by each test in
BaseDistributedSearchTestCase spinning up its own Jetty server. Initially I had
these all in a SolrCloudTestCase but I didn't notice the performance hit when
switching to the recommended BaseDistributedSearchTestCase after PR review. But
this is a really big limitation of the BaseDistributedSearchTestCase that I
hadn't thought about before. The SolrCloudTestCase sharing a cluster per class
is really a time-saver. I may move some tests which don't rely on
BaseDistributedSearchTestCase::query to LukeHandlerCloudTest as well as
"pooling assertions" to keep the number of remaining tests down in the
BaseDistributedSearchTestCase.
> Luke request handler does not know about dynamic fields on other shards fast
> enough
> -----------------------------------------------------------------------------------
>
> Key: SOLR-8127
> URL: https://issues.apache.org/jira/browse/SOLR-8127
> Project: Solr
> Issue Type: Bug
> Components: Schema and Analysis, SolrCloud
> Affects Versions: 4.10.2
> Environment: 3 shards
> Reporter: Alex K
> Assignee: Lucas Kot-Zaniewski
> Priority: Major
> Labels: dynamic, luke, pull-request-available, replication,
> sharding
> Time Spent: 5h
> Remaining Estimate: 0h
>
> Add a document with a new (never seen before) dynamic field. It will not be
> visible through Luke requests on the other shards for quite a while, and
> there is no documentation regarding exactly how long it will take. The result
> is that a query to Luke must be made to every shard in the cluster if all
> dynamic fields are needed.
> All shards should be aware of a new dynamic field within seconds, if not
> milliseconds.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
