[
https://issues.apache.org/jira/browse/SOLR-10272?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15924460#comment-15924460
]
Ishan Chattopadhyaya edited comment on SOLR-10272 at 3/14/17 3:59 PM:
----------------------------------------------------------------------
>From a user's point of view, the Collections API's CREATE command should be
>able to create a collection without specifying a configset - in that case, the
>basicconfigset could be used by default (whether this basicconfigset was
>pre-loaded when Solr started up, or the Solr process itself uploaded it from
>its install directory into ZK is an implementation detail).
Also, the Collection API's CREATE command should optionally accept a configset
(zipped), and it should create a collection based on that configset.
(Internally, it can use SOLR-6736, but that's an implementation detail.)
Also, the user could first upload a configset (using SOLR-6736) and then
specify the configName while creating a collection.
bq. I wouldn't want to lose the capability from the bin/solr script
The functionality of being able to upload a config by using bin/solr would
stay. But, the way collection creation happens now, using the bin/solr script,
could change (internally) to leverage the Solr endpoints directly (either only
the collections API or the configset API and collections API), instead of
needing to upload a configset to ZK directly.
bq. an additional collections API "BOOTSTRAPCONFIG" or something I don't see
why not. Maybe a separate JIRA?
I think SOLR-6736's configset API's upload functionality can do this.
was (Author: ichattopadhyaya):
>From a user's point of view, the Collections API's CREATE command should be
>able to create a collection without specifying a configset - in that case, the
>basicconfigset could be used by default (whether this basicconfigset was
>pre-loaded when Solr started up, or the Solr process itself uploaded it from
>its install directory into ZK is an implementation detail).
Also, the Collection API's CREATE command should optionally accept a configset
(zipped), and it should create a collection based on that configset.
(Internally, it can use SOLR-6736, but that's an implementation detail.)
Also, the user could first upload a configset (using SOLR-6736) and then
specify the configName while creating a collection.
bq. I wouldn't want to lose the capability from the bin/solr script
The functionality of being able to upload a config by using bin/solr would
stay. But, the way collection creation happens now using the bin/solr script,
could change to leverage the Solr endpoints directly, instead of needing to
upload a configset to ZK directly.
bq. an additional collections API "BOOTSTRAPCONFIG" or something I don't see
why not. Maybe a separate JIRA?
I think SOLR-6736's configset API's upload functionality can do this.
> Use a default configset and make the configName parameter optional.
> -------------------------------------------------------------------
>
> Key: SOLR-10272
> URL: https://issues.apache.org/jira/browse/SOLR-10272
> Project: Solr
> Issue Type: Improvement
> Security Level: Public(Default Security Level. Issues are Public)
> Reporter: Varun Thacker
>
> This Jira's motivation is to improve the creating a collection experience
> better for users.
> To create a collection we need to specify a configName that needs to be
> present in ZK. When a new user is starting Solr why should he worry about
> having to know about configsets before he can can create a collection.
> When you create a collection using "bin/solr create" the script uploads a
> configset and references it. This is great. We should extend this idea to API
> users as well.
> So here is the rough outline of what I think we can do here:
> 1. When you start solr , the bin script checks to see if
> "/configs/_baseConfigSet" znode is present . If not it uploads the
> "basic_configs".
> We can discuss if its the "basic_configs" or something other default config
> set.
> Also we can discuss the name for "/_baseConfigSet". Moving on though
> 2. When a user creates a collection from the API
> {{admin/collections?action=CREATE&name=gettingstarted}} here is what we do :
> Use https://cwiki.apache.org/confluence/display/solr/ConfigSets+API to copy
> over the default config set to a configset with the name of the collection
> specified.
> collection.configName can truly be an optional parameter. If its specified we
> don't need to do this step.
> 3. Have the bin scripts use this and remove the logic built in there to do
> the same thing.
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscr...@lucene.apache.org
For additional commands, e-mail: dev-h...@lucene.apache.org
