[ 
https://issues.apache.org/jira/browse/SOLR-5235?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=18132137#comment-18132137
 ] 

Nick Shanin commented on SOLR-5235:
-----------------------------------

>From code reading against apache/main b73f03976fa, the replay commit is built 
>with the REPLAY flag and sent straight to uhandler.commit, bypassing the 
>update processor chain 
>(solr/core/src/java/org/apache/solr/update/UpdateLog.java:2401-2410). The open 
>choice is whether replayed commits should go through the custom update 
>processor chain. Option one routes them through the chain so custom processors 
>see them, which changes a public plugin contract and re-runs their side 
>effects on restart. Option two keeps the direct commit and documents that 
>custom processors do not see replayed commits. Which of these two options do 
>you prefer?

🤖 AI-generated 🤖

> Update Log replay does not use the processor chain for commit
> -------------------------------------------------------------
>
>                 Key: SOLR-5235
>                 URL: https://issues.apache.org/jira/browse/SOLR-5235
>             Project: Solr
>          Issue Type: Bug
>          Components: SolrCloud
>    Affects Versions: 4.3.1, 4.4
>            Reporter: Ludovic Boutros
>            Priority: Major
>
> The update log replay sends commit commands directly with the Update Handler.
> The update processor chain is not used. I may be wrong but I think this is to 
> prevent to log this commit command again in the LogUpdateProcessor.
> But this commit command is flagged with the flag UpdateCommand.REPLAY. I 
> think this flag should be checked in the LogUpdateProcessor in order to adapt 
> its behavior.
> Currently, commit actions in custom Update Processors are not applied in case 
> of a crash without an explicit commit.
> A workaround can be done with the finish function but this is not ideal.



--
This message was sent by Atlassian Jira
(v10.3.26#10030026)

---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to