[
https://issues.apache.org/jira/browse/SOLR-9887?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15882767#comment-15882767
]
Jan Høydahl commented on SOLR-9887:
-----------------------------------
With this impl, the updated dictionary would not be stored in ZK at all, am I
right?
Isn't the workflow for the other managed filters that they update ZK then call
reload collection?
Since it is an anti pattern to keep large files in ZK I would prefer a
(generic) way to define an external source (JDBC, NoSQL, shared filesystem,
HDFS, S3, whatever) for the config, and then add REST managed API for the
component so that you can do CRUD operations on the dictionary no matter where
it resides. Not sure how you would propagate a change to all the cores in your
collection though, after having, say, added one synonym. Perhaps a reload core,
triggering each filter to consult its source and check if they have the newest
version?
So this is still a push-based approach, it could perhaps support polling to,
but I don't like the thought of 100 Solr nodes polling when it would be enough
that one did it. Do we have the concept of a collection-chief, i.e. leader of
shard 1 could by definion be the chief and the Managed Filter could have a
pollInterval setting that only the chief would act upon, and then call reload
collection when it discovers that there is an update?
> Add KeepWordFilter, StemmerOverrideFilter, StopFilterFactory, SynonymFilter
> that reads data from a JDBC source
> --------------------------------------------------------------------------------------------------------------
>
> Key: SOLR-9887
> URL: https://issues.apache.org/jira/browse/SOLR-9887
> Project: Solr
> Issue Type: Improvement
> Security Level: Public(Default Security Level. Issues are Public)
> Reporter: Tobias Kässmann
> Priority: Minor
>
> We've created some new {{FilterFactories}} that reads their stopwords or
> synonyms from a database (by a JDBC source). That enables us a easy
> management of large lists and also add the possibility to do this in other
> tools. JDBC data sources are retrieved via JNDI.
> For a easy reload of this lists we've added a {{SeacherAwareReloader}}
> abstraciton that reloads this lists on every new searcher event.
> If this is a feature that is interesting for Solr, we will create a pull
> request. All the sources are currently available here:
> https://github.com/shopping24/solr-jdbc
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
