[
https://issues.apache.org/jira/browse/SOLR-9690?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16150964#comment-16150964
]
David Smiley commented on SOLR-9690:
------------------------------------
[~michael.sun]: Wether one partitions using a shard or collection technique, I
think the end time of the lead partition (most current one) can be dynamic.
The name of the partition could be the start time. The end time of a partition
is immediately prior to whatever the start time is of the next partition. But
if there is no next partition, as in the case of the lead partition, we can
extend it as long as needed. This means we can create new partitions when the
lead partition crosses a doc count threshold.
The main challenge I see would be if you, for whatever reason, end up loading
data non-sequentially. In this event, you would be adding data to partitions
that aren't the leader, and they could grow large. You might also add data way
in the future (or simply much after the lead shard's date if even it isn't
beyond "now"), that trigger partitions to be created based on a date threshold,
yet then retroactively add data. I think this would just be a documented issue
and it's something most users could avoid or minimize.
BTW I linked this to SOLR-11299 recently.
> Date/Time DocRouter
> -------------------
>
> Key: SOLR-9690
> URL: https://issues.apache.org/jira/browse/SOLR-9690
> Project: Solr
> Issue Type: New Feature
> Security Level: Public(Default Security Level. Issues are Public)
> Reporter: David Smiley
> Attachments: DirectHashRouter.java
>
>
> This issue is for a custom Solr DocRouter that works with dates/times (or
> perhaps any field providing an int).
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
