[
https://issues.apache.org/jira/browse/SOLR-16428?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17610513#comment-17610513
]
Jason Gerlowski commented on SOLR-16428:
----------------------------------------
Tests and docs are done; I'll look to merge this soon to main, branch_9x, and
maybe branch_9_1 (as it should be a pretty safe fix to bring back).
One thing that I wanted to spell out here is the difference between the
"permissive mode" introduced in this ticket, and the sort of behavior achieved
with TolerantUpdateProcessor. The main advantage of "permissive mode" IMO is
that it allows us to be targeted/specific about the specific large-doc
exceptional case that we want to allow. In other words,
TolerantUpdateProcessor doesn't support any way to tolerate some exceptions
(the large doc case) but still be strict about others (IOExceptions, schema
issues, etc.)
> IgnoreLargeDocumentsProcessorFactory should have a "permissive" mode
> --------------------------------------------------------------------
>
> Key: SOLR-16428
> URL: https://issues.apache.org/jira/browse/SOLR-16428
> Project: Solr
> Issue Type: Improvement
> Security Level: Public(Default Security Level. Issues are Public)
> Components: UpdateRequestProcessors
> Affects Versions: 9.0, main (10.0)
> Reporter: Jason Gerlowski
> Assignee: Jason Gerlowski
> Priority: Minor
> Time Spent: 1h 10m
> Remaining Estimate: 0h
>
> IgnoreLargeDocumentProcessorFactory only has a single way to handle documents
> that exceed its configurable size limit. The first violation throws a
> SolrException: in effect, short-circuiting any remaining documents in the
> "batch" and returning a 400 to the user.
> This is great for end users whose clients are built to handle the resulting
> 400 response, and who can modify and resubmit the batch. But it's not ideal
> for every use-case, especially where "best-effort" indexing is good enough.
> This ticket proposes adding a new "permissive" mode of handling too-large
> documents to ILDPF. Under this new mode "too-large" documents will be logged
> (and not indexed), but won't cause the entire batch to be aborted/error-out.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
